Recreate Our 1-Minute Seedance 4K Film - The Full Seedance 4K Breakdown With All Prompts 重现我们那部 1 分钟的 Seedance 4K 短片——Seedance 4K 全流程解析及所有提示词

Recreate Our 1-Minute Seedance 4K Film -  The Full Seedance 4K Breakdown With All Prompts

Seedance 2.0 just came out in 4K, so we used it to make a hyper-realistic one-minute short film - and this is the full breakdown. Every scene, every asset, every prompt.Seedance 2.0 刚发布了 4K 版本,我们用它做了一部超写实的一分钟短片。这是完整的解析,包括每一个场景、每一个素材以及每一个提示词。

The whole film runs on the simplest workflow there is: Claude writes the prompts, Higgsfield generates the content. For every scene you need three ingredients - the props, the characters, and the location. Build those as assets first, and the video generation almost takes care of itself.整部片子用的流程很简单:Claude 写提示词,Higgsfield 生成内容。每个场景都需要三个要素——道具、角色和场景。先把这些做成素材,视频生成起来基本就不用操心了。

The idea of the film: a guy sits on the couch flipping channels, completely glued to the hyper-realistic 4K footage on the screen, while his mom keeps calling him for dinner.短片的构思是这样的:一个男的坐在沙发上换台,被屏幕里超写实的 4K 画面迷住了,他妈在旁边一直喊他去吃饭。

In this post you'll find the exact prompts, the generations, and the Claude prompt-writing skill we use for everything. Copy any prompt, hit Recreate, and build your own..这篇文章里有确切的提示词、生成结果,还有我们用来处理一切的 Claude 提示词技巧。你可以复制任何提示词,点击“重现”,然后做出你自己的作品。


Scene 1 — The character, the mom, the room, and the remote场景 1 —— 角色、妈妈、房间和遥控器

For the first scene we need four assets: the main character, his mom, the living room, and a TV remote as a prop.第一个场景需要四个素材:主角、他妈、客厅,还有一个遥控器作为道具。

The character (new outfit)角色(新装束)

1

We already had a character sheet, so we just asked Claude for a prompt to change the outfit — a Y2K look with a pink "HIGGS" tee and a teddy bear on the belt.我们已经有角色设定图了,就让 Claude 写了个换装提示词——千禧风(Y2K),穿粉色“HIGGS”T恤,腰带上挂个泰迪熊。

1 2
SUBJECT (RIGHT panel — large head): @image_1 — face matches input 100%. Large head-and-shoulders close-up, neutral resting expression, calm direct gaze, same identity / features / hair / skin as the reference, no drift. Across the bridge of the nose: a stylish athletic nose strip with a small floral print — delicate flowers on the tape, clean and fashion-forward, sitting flat on the skin. Wearing the SAME outfit as the LEFT panel — upper body visible: washed raspberry-pink retro tee with pale-grey contrast collar trim, the top edge of the chrome-bubble "HIGGS" chest graphic just visible. Accessories on the head/neck only: small silver hoop in the left ear plus a second tiny stud, thin delicate silver chain at the collar layered with a slim beaded cord. Hands and arms NOT in frame — clean head-and-shoulders crop, no hands near the face or shoulders. LEFT panel — OUTFIT ONLY, ghost-mannequin (invisible body, garments floating/styled as worn, no person, no head): Vintage-inspired Y2K casual streetwear sized for a ~20-year-old, ~180cm frame. Hero piece: washed raspberry-pink retro tee, relaxed boxy cut, sun-faded cotton with a worn texture, mock layered look with a contrast pale-grey trim at the collar and short sleeve cuffs. Across the chest: a stylish Y2K-style "HIGGS" wordmark — chrome-bubble lettering with a soft metallic sheen and subtle gradient, slight italic motion, early-2000s graphic feel, sitting cleanly in the fabric weave (not glossy sticker). Bottoms: oversized wide-leg carry/cargo-style trousers in faded khaki-tan, heavy workwear cotton, dropped crotch, curved panel seams, full length — long hem stacking and puddling over the shoes, no cuff, no roll-up, ankles fully covered. A bright blue woven belt threaded through the loops. Hanging from a belt loop on a clip: a small brown plush teddy-bear charm. Footwear: low-profile suede sneakers in dusty grey-brown with gum soles, mostly hidden under the long trouser hem. Accessories: a couple of stacked silver rings, a slim wrist cuff. Overall vibe: thrifted retro, soft nostalgic palette, playful and lived-in. LAYOUT: Character-sheet split composition. LEFT = full outfit in WIDE shot, ghost-mannequin style, no body. RIGHT = large head-and-shoulders close-up. Clear vertical divide. BACKGROUND + LIGHT: Seamless neutral grey studio backdrop. Even soft neutral studio light, balanced exposure, no harsh shadows, ~5500K neutral white balance. STYLE: Clean studio character-reference sheet, photoreal, sharp detail. Natural skin tones, fabrics read as real cotton/nylon with weave and wear.主体(右侧面板——大头像): @image_1 — 面部与输入图 100% 匹配。大头像特写,表情平静自然,目光直视,身份、特征、发型、肤色与参考图一致,无偏差。鼻梁上贴着时尚的运动鼻贴,上面有细碎的花卉图案——花纹精致,平整贴合皮肤。穿着与左侧面板相同的装束——露出上半身:水洗覆盆子粉色复古 T 恤,浅灰色对比领口,铬合金泡泡字体“HIGGS”胸前印花清晰可见。头颈部饰品:左耳戴小银圈耳环及一个细小的耳钉,领口处戴着细银链,叠戴一根细珠绳。手和手臂不在画面内——干净的头像裁切,手部不靠近脸部或肩膀。 左侧面板——仅装束,幽灵模特(隐形身体,衣物悬浮/展示穿着效果,无人,无头): 适合 20 岁左右、180cm 身高的千禧风复古休闲街头装。核心单品:水洗覆盆子粉色复古 T 恤,宽松箱型剪裁,水洗棉质感,假两件领口设计,浅灰色领口和袖口滚边。胸前:时尚的千禧风“HIGGS”字样——铬合金泡泡字体,带有柔和的金属光泽和细微渐变,略带动态感,2000 年代初期图形感,印花自然融入布料纹理(非胶印质感)。下装:褪色卡其棕色阔腿工装裤,重磅工装棉布,低裆,弧形拼接缝,全长——裤脚堆叠在鞋面上,无卷边。蓝色的编织腰带。腰带扣上挂着一个棕色泰迪熊玩偶。鞋子:灰棕色绒面革低帮运动鞋,大部分被长裤遮住。配饰:叠戴银戒指,细手环。整体风格:复古旧货,怀旧色调,随性自然。 布局: 角色设定表分割构图。左侧 = 宽幅全身装束,幽灵模特风格,无身体。右侧 = 大头像特写。中间有清晰的垂直分割线。 背景与光线: 无缝中性灰摄影棚背景。均匀柔和的摄影棚光,曝光平衡,无生硬阴影,~5500K 中性白平衡。 风格: 干净的摄影棚角色参考表,照片级真实感,细节锐利。肤色自然,布料呈现真实的棉/尼龙织物纹理和磨损感。

We ran it in GPT Image 2 at 4K quality - the better the input image, the sharper the final video.我们在 GPT Image 2 上用了 4K 画质运行——输入图越好,最终视频越清晰。

Two things matter here. First, the grey background: after tons of testing, character sheets on grey perform way better than on white or black. Second, make sure you have one image where the face is clearly visible and one full-body for the outfit - that way Seedance doesn't have to guess anything.这里有两点很重要。第一,灰色背景:经过大量测试,灰色背景上的角色表效果远好于白色或黑色。第二,确保有一张脸部清晰的图和一张全身装束图——这样 Seedance 就不用瞎猜了。

Then click "Create Element," name it, and save. Do this for every character, location, and prop you make — Claude will automatically reference these images in the prompts.然后点击“创建元素”(Create Element),命名并保存。为你制作的每一个角色、场景和道具都这样做——Claude 会在提示词中自动引用这些图片。

The mom妈妈

We attached the character sheet to Claude and asked for a mom aged 40-45 in casual home clothes who clearly resembles her son - 3 views on the same grey background.我们将角色设定图发给 Claude,要求生成一个 40-45 岁、穿着居家服、长相明显像她儿子的妈妈——在同一个灰色背景下展示 3 个视角。

1 3
SCENE CONTEXT A character reference for the mother of the young man in reference 1. She must read instantly as his mother — same heritage and facial DNA, older generation — dressed in comfortable home clothes for cooking (wardrobe per references 2, 3, 4). All three shots are taken in a clean studio on a neutral light-grey cyclorama: a tight face portrait, a full-length front view, and a full-length back view. ACTIVE REFERENCES Ref 1 = son's face (identity/resemblance source). Refs 2, 3, 4 = homewear cooking wardrobe source (clothing only, NOT the kitchen environment). CHARACTER ANCHORS (identity lock — family resemblance to the son) Woman, around 50–55. Warm olive-tan skin with the same undertone as the son, mixed Latino–Southeast-Asian heritage. Dark hair with natural curl and a few soft grey strands, mid-length, loosely pulled back with curls framing the face. Inherited facial structure straight from the son: warm dark-brown almond-shaped eyes with the same spacing and slight downturn, the same broad nose with a rounded tip, full lips, soft rounded cheeks, strong defined dark brows, the same calm steady gaze. The resemblance must be unmistakable — clearly mother and son. A neat, fine, well-groomed thin mustache along the upper lip — light, sparse, tidy and intentional, NOT thick, NOT bushy or voluminous. Small silver hoop earring and a tiny stud, like the son. Warm composed expression, neutral lips. Realistic skin texture, age-appropriate: natural lived-in skin with visible fine pores and soft sheen, a few light horizontal forehead lines and fine wrinkles under the eyes, plus faint soft smile lines. The wrinkles read clearly but stay shallow and gentle — fine and natural, NOT deep, NOT heavy or harsh. Minimal makeup, real human skin, no plastic or over-smoothed retouching. 100% consistent across all three shots. WARDROBE (homewear for cooking, per refs 2–4) Soft, relaxed domestic clothing in warm natural tones: a light linen or cotton camp-collar shirt or a white long-sleeve tee with sleeves rolled to the forearm, layered under a mid-length cotton/linen kitchen apron in beige, clay, or muted olive, tied at the waist with soft strings. Relaxed comfortable trousers below in a warm neutral tone. Soft, lived-in, gently creased fabrics. Consistent outfit across the full-length front and back shots. SHOT 1 — FACE PORTRAIT (close-up) Framing: tight close-up, head and shoulders, face fully readable and centered. Optics: 29° diagonal field of view, short telephoto portrait lens character, camera 4 m back, flattering stable proportions, subject razor-sharp, skin texture and fine wrinkles clearly resolved. Camera: eye-level, locked static. SHOT 2 — FULL-LENGTH FRONT Framing: full figure head to feet, standing relaxed and facing camera, arms natural at sides, the home cooking outfit fully readable. Optics: 47° standard normal lens character, camera 4–5 m back, zero distortion, straight rectilinear edges, deep focus head to hem. Camera: eye-level, locked static, full body centered with even margins. SHOT 3 — FULL-LENGTH BACK Framing: full figure from directly behind, same standing pose, showing the back of the shirt, apron ties, hair, and trousers. Optics: same 47° standard normal lens, same distance and height as Shot 2 for matched scale. Camera: eye-level, locked static, dead-on rear view. BACKGROUND All three shots on a clean, seamless neutral light-grey cyclorama — smooth infinity-wall studio sweep with no visible corners, floor seam, or horizon line. No kitchen, no props, no environment. Background stays flat neutral grey across all three frames. LIGHTING Even soft studio softbox key from the front-upper, broad and gentle, light wraparound fill, smooth shadowless gradient on the grey cyclorama, soft contact shadow at the feet in the full-length shots. For the portrait, the same soft key so the face, fine mustache and shallow wrinkles read clearly without harsh shadow. Natural warm skin rendering, no harsh speculars. PHYSICS & MATERIAL Realistic human skin with fine pores, soft sheen, and shallow natural wrinkles on the forehead and under the eyes. Soft linen and cotton with natural folds and light creases, apron fabric draping with real weight, hair holding natural curl with slight movement, jewelry sitting with weight. Grounded, photoreal studio texture, no over-smoothing. POSITIVE LOCKS The woman is clearly the mother of the young man in ref 1 — shared olive-tan skin, dark curly hair, brown almond eyes, broad rounded nose, and family likeness make the relationship obvious. Mustache stays fine, neat and well-groomed, never thick or bushy. Skin is realistic with shallow, non-deep forehead and under-eye wrinkles. Same face, hair and wardrobe consistent across all three images (outfit matched in the two full-length shots). Neutral light-grey cyclorama studio background in every shot. Single subject, no extra people, no text overlays, no watermark. Photoreal editorial studio photography.

Pick the one where the details look most natural, save it as an element, and move on.挑一张细节最自然的,保存为元素,然后继续。

The location场景

1 4
Modern open living hall with a teenage vibe, 3/4 angle wide shot. Spacious room that doubles as a teen's hangout — lived-in and a little messy, not a showroom. Large modern flat-screen TV (~65") mounted on the main wall as the focal point, screen off/black, game controller on the low media console below it. Comfortable sofa with cushions dented and askew, a crumpled throw blanket half-falling off, a beanbag chair on the floor. Hoodie and a couple of t-shirts tossed around, sneakers kicked off near the rug, headphones slung over the sofa arm, a few cups and a snack wrapper on the console, charging cable trailing on the floor. String fairy lights and an LED strip glow along the wall, a couple of plain framed posters (blank, no text or images), soft area rug slightly bunched, faint scuffs on the wall. NO computer desk. Generic unbranded objects — NO logos, NO brand names, NO visible text anywhere. Natural window daylight + soft warm lamp. Photorealistic, 2020–2022 interior style. 16:9.现代开放式客厅,带有青少年氛围,3/4 角度广角镜头。宽敞的房间,兼作青少年的休闲区——有生活气息,稍微有点乱,不是样板间。主墙上挂着一台大型现代平板电视(约 65 英寸)作为视觉焦点,屏幕关闭/黑色,下方的矮媒体柜上放着游戏手柄。舒适的沙发,靠垫凹陷歪斜,一条皱巴巴的毯子半掉在地上,地上有个懒人沙发。到处扔着连帽衫和几件 T 恤,地毯旁踢着运动鞋,沙发扶手上挂着耳机,柜子上放着几个杯子和零食包装袋,充电线拖在地上。墙上挂着串灯和 LED 灯带,几张普通的带框海报(空白,无文字或图像),地毯稍微有些起皱,墙上有淡淡的擦痕。没有电脑桌。通用无品牌物品——没有 Logo,没有品牌名,画面任何地方都没有可见文字。自然窗光 + 柔和暖灯。照片级真实感,2020-2022 年室内风格。16:9。

The key word in this prompt is the 3/4 angle - it shows more of the room, so Seedance gets better depth perception and renders it without breaking anything. For locations we always use Soul Cinema: it's built for cinematic shots. Your video quality depends mostly on this one image - examine every detail, and if something looks off, regenerate.这个提示词的关键在于 3/4 角度——它能展示更多房间细节,所以 Seedance 能获得更好的深度感知,渲染时也不会出错。对于场景,我们总是用 Soul Cinema:它是专门为电影级镜头设计的。视频质量主要取决于这张底图——仔细检查每一个细节,如果看起来不对劲,就重新生成。

The remote (prop) 遥控器(道具)

1 5
SCENE CONTEXT A single modern 4K Smart TV remote control prop, stripped down to a minimal button set, presented as a clean three-view product reference: front face, true side profile, and back shell. No hands, no environment, object only on a flat neutral medium-grey seamless background. OBJECT Slim minimalist Smart TV remote, contemporary 2020s design language, matte-black soft-touch polymer body, rounded corners, thin tapered thickness. Reduced, primitive button layout — no streaming-app shortcut buttons (no Netflix, Prime Video, YouTube, Disney), no mic button, no back/home buttons. Front face layout, top to bottom: A single round power button at the very top-right. Directly below the power area, a "higgs 4K" wordmark in bright lime-chartreuse green, hex #D2F84E — sized slightly smaller than a hero logo, compact and tidy in the upper face, not oversized. Below the logo, a large central circular navigation ring with a glossy black round OK/select key in the middle and a thin chrome ring; four directional arrows (up, down, left, right) printed on the ring at top, bottom, left and right positions. Below the ring, two simple control pairs side by side: a volume rocker on the left (a "+" key above a "−" key, small "VOL" label) and a channel rocker on the right (an up-chevron above a down-chevron, small "CH" label). Clean, uncluttered, lots of empty matte-black space, premium minimalist aesthetic. Back shell: smooth tapered surface with a subtle recessed battery-compartment cover and a small "higgs 4K" logo in lime-chartreuse #D2F84E near the bottom. LAYOUT Three identical remotes arranged left to right on the same horizontal baseline, evenly spaced, equal scale. Left = front view, face toward camera, fully readable. Center = strict side profile, edge-on in plane, showing thickness and slightly raised buttons in silhouette. Right = back view, rear shell toward camera. Object centered vertically, generous equal margins, no overlap, no cropping. OPTICS 47° diagonal field of view, standard normal lens character, camera 3 to 4 meters back, zero distortion, straight rectilinear edges, natural proportions, deep depth of field so all three views stay razor-sharp front to back. CAMERA Eye-level, locked static, dead-on orthographic framing for each view. No perspective tilt, no foreshortening. LIGHTING Even soft studio softbox key from the front-upper, broad and shadowless, subtle gradient falloff on the grey background. Soft contact shadow directly beneath each remote to anchor it. No harsh reflections, no blown speculars, matte surfaces read as matte, the lime logo stays saturated and clean. MATERIAL Soft-touch polymer absorbs light, rubber buttons show faint sheen, the lime-chartreuse "higgs 4K" logo reads as crisp colored print, chrome navigation ring shows a thin clean highlight. Pristine, dust-free, no fingerprints. POSITIVE LOCKS Minimal primitive 4K Smart TV remote only — buttons limited to: power, central directional ring with OK and four arrows, volume rocker, channel rocker. No app shortcut buttons, no mic, no back/home keys. No hands, no props, no text overlays except the "higgs 4K" logo and the small VOL/CH labels and arrow glyphs. The "higgs 4K" logo is slightly smaller than the current version, compact and balanced. Logo color stays exactly lime-chartreuse hex #D2F84E on both front and back. Background stays flat neutral medium grey. All three views share identical body, color, scale, and lighting. Logo spelling stays exactly "higgs 4K".

Treat prop sheets the same way as character sheets: grey background, multiple views.道具表的处理方式和角色表一样:灰色背景,多视角。

Here's a real lesson from this step: the GPT Image 2 results looked flat - no depth, no shadows, and Seedance might render that as a 2D object. We switched to Nano Banana Pro with the same prompt, and the remote got depth, highlights, and realistic tapered edges. There's no single best model for everything - if the result isn't what you wanted, switch the model or tweak the prompt.这一步有个教训:GPT Image 2 的结果看起来很平——没有深度,没有阴影,Seedance 可能会把它渲染成 2D 物体。我们改用了 Nano Banana Pro,同样的提示词,遥控器就有了深度、高光和真实的边缘。没有哪种模型是万能的——如果结果不理想,就换个模型或者改改提示词。

The first video第一个视频

With all assets ready, we uploaded them to Claude and described the shot: the character walks into the room, sits on the couch, and turns on the TV - one continuous handheld take with optical flares.素材准备好后,我们把它们上传给 Claude 并描述镜头:角色走进房间,坐在沙发上,打开电视——一个连续的手持镜头,带有光晕效果。

SCENE CONTEXT Evening. @ADIL comes home, crosses his living room, sits on the sofa facing the TV, picks up the remote and switches it on — one unbroken take. From the unseen kitchen his mother calls in English that food is almost ready; he answers without taking his eyes off the screen. ACTIVE REFERENCES @ADIL — young man, tired end-of-day, settling into his own space. 100% matches the reference. @REMOTE — ordinary household TV remote. 100% matches the reference. (Mother is an offscreen English voice only — no in-frame reference, never enters the shot.) LOCATION MAP @ADIL'S-ROOM — STYLE REFERENCE ONLY, not a fixed keyframe; model extends the world, the subject moves through real space, never pinned 1:1. Foreground: doorway / hallway frame-right leading off to an unseen kitchen. Midground: the sofa, facing forward. Background: the far wall with the flatscreen TV. A side window sits frame-left as the light source. Lived-in room, soft daylight, dust and faint haze in the air. FIRST FRAME / BLOCKING Non-empty opening frame: @ADIL caught mid-step in the doorway, three-quarter back to camera, hand brushing the frame, the room opening up ahead of him with the TV visible on the far wall directly forward. Sofa in the midground between him and the TV. Asymmetric framing, subject already in motion. FORMAT MODE One continuous shot — the camera does not cut on its own. No edits, no transitions. OPTICS Continuous evolving reframe inside the single take: opens MS at 47° from three-quarter behind, arcs around to the front and eases to a waist-level MS on the remote, then settles into an MCU at 29° on hands and face as the TV ignites. Rectilinear with anamorphic optical flares, prime-lens character, 180° shutter motion blur. CAMERA One continuous handheld take, live operator breathing, organic shake and micro-drift, a subtle dutch that resettles. Eye-level following him in from three-quarter behind as window flares streak past, arcing to the front as he sits, dropping to a slightly low waist-level reframe onto @REMOTE in his hand as he answers his mother, then drifting tighter onto hands and face for the click. Focus rides with him. Camera on the shadow side; operator axis travels from behind to front. The lens never stops moving — one unbroken move motivated by following him home and landing on the click. ACTION Subject: steps in through the doorway, eyes scanning, unhurried walk toward the sofa, shoulders dropping; sits down facing the TV with a settling exhale; tips his head slightly toward the hallway as his mother calls, answers back without looking away, hand reaching and closing around @REMOTE on the cushion; lifts @REMOTE, aims it forward at the TV, a half-beat pause, thumb presses — soft click — the screen ignites and cool glow washes back across his face from ahead. Camera: described in CAMERA — one continuous following move, behind to front to medium-close, no self-cuts. PERFORMANCE Tired end-of-day release: a slow exhale with visible chest rise as he sits, shoulders loosening, jaw unclenching. Eyes drift low, then catch the screen light — a small flicker of anticipation, lips parting slightly. Pore-level realism: vellus hair, capillary flush, wet living eyes with catch-lights from the screen. PHYSICS @REMOTE carries real weight in his hand. The cushion compresses and the fabric shifts as he sits. Correct contact shadows under him on the sofa and on the floor. Nothing floats. LIGHTING Natural light only. Soft window daylight key from frame-left, camera on the shadow side, atmospheric haze carrying the optical flares at roughly 10–15 meters depth. As the TV ignites, its cool screen glow becomes a second key from directly in front, washing his face from ahead. WB locked 5200K. COLOR GRADE Warm lived-in room neutrals as the dominant material, cool window light and TV glow as the secondary beam, and a chartreuse #d1ef17 accent living only in the TV standby LED, the first flicker of the screen, and a single glint catching a lens flare — color tied to source and surface, not a flat list. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. Room ambience, footsteps on the floor, the soft compression of the cushion, a soft remote click, the low power-on hum of the TV. Dialogue (English, audio only, no subtitles): Mother (offscreen, from the kitchen): "Food's almost ready, sweetie!" @ADIL (answering, eyes on the screen): "Okay, Mom — coming soon!" STYLE 8K photorealistic — no 3D render, no game engine, no game-cutscene aesthetic. Naturalistic master cinematography, anamorphic optical flares, fine film grain. OUTPUT SETTINGS 16:9. Real-time throughout, no slow-motion. POSITIVE LOCKS One continuous shot, the camera does not cut on its own. The TV stays on the far wall directly in front of the sofa and in @ADIL's eyeline — he faces it, screen glow comes from ahead of him. @REMOTE stays normal household remote size. The mother stays offscreen, an English voice only. TV screen content stays abstract and unbranded — plain glow, no logos, no UI, no subtitles. Contact shadows on sofa and floor read clearly. Wardrobe comes from the @ADIL reference and stays identical across the whole take. Eyes stay natural — no eye glow.

Claude returns the full prompt with the assets already locked in - that's the prompt-writing skill doing the work. Download the skill here →Claude 返回了锁定了素材的完整提示词——这就是提示词编写技巧的作用。点击这里下载这个技能 →

prompt-writter.skillprompt-writter.skillapplication/zipprompt-writter.skillapplication/zip

Scene 2 - The fantasy flight on the TV场景 2 —— 电视里的奇幻飞行

Now the shot that plays on the TV: a girl flying on a mystical creature through a jungle of giant trees.现在是电视里的镜头:一个女孩骑着神秘生物在巨树丛林中飞行。

2
Photorealistic cinematic creature reference, a colossal mystical jungle dragon-bird hybrid mount, half majestic dragon and half tropical bird, film still from a high-budget fantasy movie, shot on ARRI Alexa, 85mm lens, VFX quality, hyper-detailed, 8K, dramatic realism. A large graceful rideable creature built for flight through dense jungle. Heron-like elongated neck and a sharp dagger beak opening into a faint reptilian maw with a glowing amber gape; small backswept horns and a vivid spined feathered crest; ancient luminous golden eye. Plumage in lush emerald green, jade, turquoise and gold, blending into deep teal flight feathers and a pale cream throat — iridescent like tropical birds. Legs, wing joints and tail base transition into fine iridescent green-gold dragon scales. Huge powerful wings: feathered like a bird but with subtle clawed dragon finger-bones at the leading edge, broad for agile jungle flight. Strong legs with gripping talons for perching on giant branches, a long flowing tail mixing feathered streamers with spined scaled ridges for steering. Natural saddle-like hollow between the shoulder blades with two raised crest ridges that act as handholds, simple woven-vine and leather riding harness with bronze fittings strapped across the chest and neck — clearly a mount a person can ride. Faint bioluminescent green energy and drifting jungle mist curl around the body. Every surface physically real: feather barbs and keratin, iridescent scales with subsurface sheen, realistic light scatter, micro-detail, photographic depth. Awe-inspiring scale, agile and noble — equal parts tropical bird and ancient jungle dragon. Reference sheet layout on a seamless neutral gray backdrop (#7f7f7f), soft diffused studio lighting: full-body side profile, front view, three-quarter view, head close-up, wing-spread detail, tail detail, and saddle/harness detail. Negative: pure dragon, pure bird, cartoon, flat illustration, painterly brushstrokes, low detail, plastic CGI, oversaturated, blurry, watermark, text, distorted anatomy.照片级电影感生物参考,巨大的神秘丛林龙鸟混种坐骑,一半是威严的龙,一半是热带鸟,高预算奇幻电影剧照,ARRI Alexa 拍摄,85mm 镜头,VFX 质量,超细节,8K,戏剧性真实感。 一只巨大的、优雅的、可骑乘的生物,专为在茂密丛林中飞行而生。苍鹭般的长颈,尖锐的匕首状喙张开,露出带有微弱爬行动物特征的喉咙,喉咙内部发出微弱的琥珀色光;头部有向后弯曲的小角和生动的羽毛冠;古老的金色发光眼。羽毛呈翠绿色、翡翠色、绿松石色和金色,过渡到深青色的飞羽和浅奶油色的喉咙——像热带鸟类一样具有虹彩。腿部、翼关节和尾部基部过渡到精致的虹彩绿金色龙鳞。巨大的强力翅膀:羽毛像鸟,但前缘有隐约可见的带爪龙指骨,宽大适合在丛林中敏捷飞行。强壮的腿部带有抓握力强的爪子,用于栖息在巨大的树枝上,长长的尾巴混合了羽毛飘带和带刺的鳞片脊,用于转向。 肩胛骨之间有一个天然的鞍状凹槽,有两个隆起的脊作为把手,简单的藤蔓和皮革编织的骑乘鞍具,带有青铜配件,横跨胸部和颈部——显然是人可以骑乘的坐骑。微弱的生物发光绿色能量和漂浮的丛林薄雾围绕着身体。 每一个表面都物理真实:羽毛倒钩和角蛋白,带有次表面光泽的虹彩鳞片,真实的光散射,微细节,摄影深度。令人敬畏的体型,敏捷而高贵——既是热带鸟类,也是古老的丛林龙。 无缝中性灰背景(#7f7f7f)上的参考表布局,柔和的漫射摄影棚光:全身侧面、正面、3/4 视角、头部特写、翅膀展开细节、尾部细节以及鞍具/挽具细节。 负面提示:纯龙,纯鸟,卡通,平面插画,绘画笔触,低细节,塑料感 CGI,过度饱和,模糊,水印,文字,解剖结构扭曲。
2
Photorealistic cinematic character reference of a young woman as a forest-spirit nymph, film still quality, shot on ARRI Alexa, 85mm lens, natural realism, hyper-detailed, 8K. A natural, real-looking woman in her mid-20s with a warm, serene expression. Long flowing golden-blonde wavy hair cascading past her shoulders, adorned with a lush crown of real wildflowers (purple, yellow, white blooms), green leaves and small buds woven throughout. Soft natural face with sun-touched skin, light freckles, expressive hazel eyes, subtle natural look. Wearing an organic forest outfit made of nature: an off-shoulder cropped top of woven moss, soft bark fibers and dried grasses; a layered skirt of overlapping fresh green leaves, twisted vines and frayed natural fibers, decorated with raw crystals and polished gemstones (clear quartz, turquoise, amber, opal) hanging like jewelry. Vine-wrapped bracelets and armbands, a beaded gemstone necklace with a crystal pendant, small flowers tucked throughout. Barefoot. Earthy, handcrafted, living-forest look — no weapons, nothing in her hands. Every detail physically real: skin pores and texture, individual hair strands, real flower petals and leaves, woven fiber and vine texture, crystal refractions, realistic soft forest daylight. Believable and grounded. Reference sheet layout on a seamless neutral gray backdrop (#7f7f7f), soft diffused studio lighting, exactly three views: full-body front view, full-body back view, and one large close-up of the face. Negative: weapons, staff, stick, cartoon, illustration, painterly, plastic CGI, heavy makeup, oversaturated, blurry, watermark, text, distorted anatomy, extra fingers.照片级电影感角色参考,年轻女性森林精灵,电影剧照质量,ARRI Alexa 拍摄,85mm 镜头,自然真实感,超细节,8K。 自然、真实的 20 多岁女性,表情温暖宁静。长长的金色波浪卷发披肩,戴着由真正的野花(紫色、黄色、白色花朵)、绿叶和小花蕾编织而成的花环。柔和的自然面部,皮肤有阳光亲吻的痕迹,淡淡的雀斑,富有表现力的榛色眼睛,自然的妆容。 穿着由自然材料制成的森林装束:露肩短上衣,由编织的苔藓、软树皮纤维和干草制成;叠层裙子由重叠的鲜绿叶子、扭曲的藤蔓和磨损的天然纤维组成,装饰着像珠宝一样悬挂的原始水晶和抛光宝石(透明石英、绿松石、琥珀、蛋白石)。藤蔓包裹的手镯和臂环,带有水晶吊坠的珠串宝石项链,全身点缀着小花。赤脚。质朴、手工制作、活生生的森林外观——没有武器,手中没有东西。 每一个细节都物理真实:皮肤毛孔和纹理,单根发丝,真正的花瓣和叶子,编织纤维和藤蔓纹理,水晶折射,真实的柔和森林日光。可信且扎实。 无缝中性灰背景(#7f7f7f)上的参考表布局,柔和的漫射摄影棚光,正好三个视角:全身正面、全身背面,以及一张大的面部特写。 负面提示:武器,法杖,棍子,卡通,插画,绘画感,塑料感 CGI,浓妆,过度饱和,模糊,水印,文字,解剖结构扭曲,多余的手指。

For characters from scratch, Soul Cinema is the best and the cheapest - 1 credit for 8 images对于从零开始的角色,Soul Cinema 是最好也最便宜的——1 个积分生成 8 张图。

SCENE CONTEXT @woman flies on @fantasy-dragon, a large flying creature, through a colossal jungle of mile-high fantastical trees. Wordless throughout, she steers with her body and pure joy — laughing and whooping as @fantasy-dragon pulls off daring tricks: a sheer near-vertical plunge down an immensely tall 90° waterfall almost grazing the water, a low skim across the pool, passing through a cloud of giant butterflies with a single breathtaking slow-motion moment, gliding between branches and lianas past leaping monkeys, threading a razor-narrow gap, then climbing straight up the trunks and bursting above the canopy into a vast reveal of the whole world. One continuous soaring flight. ACTIVE REFERENCES @fantasy-dragon: the flying creature. 100% matches the reference, identical design every frame. @woman: the rider, seated just behind @fantasy-dragon's shoulders, hands on its shoulder-ridge, leaning, ducking and bracing into every bank, dive and roll. 100% matches the reference, same person throughout. CREATURE SCALE & WING PHYSICS (behavior only, design from reference) @fantasy-dragon is large — wingspan of a small aircraft, body big enough to carry @woman easily, powerful and substantial. Its wings obey real flight physics with three clear modes: (1) in the vertical dive, wings tuck in tight and folded against the body, streamlined for a fast nose-down plunge; (2) when gliding and weaving through the trees, wings hold wide, spread and still, soaring on locked wings — steering by tilting and angling the spread wings and body, not by flapping; (3) when climbing and gaining height, broad, deliberate, weighty wingbeats with a clear up- and down-stroke through a large arc, the membrane flexing and the whole body rising and dipping with each beat. Mass and inertia read in every motion. FORMAT MODE Single continuous take with no cuts. One in-camera speed ramp: real-time motion throughout, ramping down smoothly into super slow motion for a single butterfly close-up around the 4–6s mark, then snapping back hard and instantly to full real-time speed. Real-time everywhere else. 15 seconds, 16:9 anamorphic widescreen. The clip is purely visual and sound-effect driven, with a fully music-free, speech-free, caption-free track — natural ambience and the rider's wordless voice only. CAMERA POSITION LOCK — ALWAYS BEHIND The camera stays locked behind @fantasy-dragon and @woman for the entire clip — a chase camera that follows directly from the rear, holding their backs in frame with the world rushing toward and past them. We always see them from behind: the back of @woman's head and shoulders, @fantasy-dragon's wings filling the foreground, the jungle opening up ahead. During the slow-motion beat the camera keeps its rear position and simply racks focus onto a butterfly hovering close to the lens, with @fantasy-dragon and @woman gliding softly just beyond it. The camera rides their tail through the dive, skim, glide, gap and climb, keeping the same rear vantage throughout. THE LOCATION (full description — maximum color and life) A surreal, monumental rainforest of impossible scale. The trees are mile-high fantastical giants — colossal pillar-trunks wider than towers, bark in deep browns and mossy greens, crowns vanishing into golden mist; upper branches fuse into vaulted green cathedral arches. Thick horizontal branches and long hanging lianas cross the flight path at many depths, and one pair of trunks leans together to form a razor-narrow vertical gap. A colossal, immensely tall waterfall drops in a sheer, near-vertical 90° wall — a vast curtain of falling water plunging straight down an enormous cliff-trunk from high in the canopy, falling and falling all the way to the forest floor, into a glowing turquoise pool and mist-filled gorge, sunlight throwing rainbows across the spray. The whole jungle blazes with color and life: cascading flowers in magenta, violet, orange and gold, glowing bioluminescent blooms, hanging curtains of pink and blue blossom, giant lily-pads and lotus on the pool, emerald spiral ferns, turquoise moss, dangling vines heavy with fruit, oversized jewel-toned leaves, drifting seed-pods and floating spores catching the light. It teems with diverse wildlife at every depth — dense flocks of brilliant birds (scarlet macaw-like fliers, cobalt songbirds, gold-and-green parrots, iridescent hummingbirds, big slow herons over the water), vast drifting swarms of giant luminous butterflies in sapphire, amber and emerald, dragonflies and glowing fireflies, troops of long-limbed monkeys swinging and leaping from tree to tree, neon tree-frogs and chameleons on the trunks, a coiled serpent on a branch, a pair of huge gentle deer-like grazers and a family of small spotted creatures drinking at the pool far below. High screen-right, warm golden sun rakes down through the canopy in thick volumetric god-rays, catching waterfall-mist, pollen and floating spores. Everything alive is tiny against the trees. Deep parallax: foreground vines, flowers and leaves rush past, midground trunks and wildlife slide by, the far canopy and sky shift slowly behind. FIRST FRAME AND SPATIAL BLOCKING The first visible frame already contains @fantasy-dragon and @woman, seen from behind, hanging high at the very top of the immensely tall 90° waterfall — tiny against the towering wall of water, already tipping nose-down into the vertical plunge with wings folding in. The camera sits directly behind them. @fantasy-dragon center frame, x 50%, y 40%, wings tucking tight, pointing straight down the falling water; @woman just ahead of the lens on its back, leaning low and forward into the dive, the back of her head and streaming hair in frame, laughing. The flight path drops straight down the huge waterfall almost touching its surface, flares open just above the pool, skims the water, rises into the butterflies and the slow-motion beat, glides through the trees, threads the narrow gap, then climbs upward — always viewed from the rear. OPTICS 107° diagonal field of view, wide rectilinear lens character, camera held very close — 0.5–0.8 m behind @fantasy-dragon's back and wings, almost riding with them. Immediate foreground (@fantasy-dragon's wings and body, @woman's streaming hair, rushing spray, vines and butterflies) looms large; the surrounding jungle spreads wide to all frame edges; straight trunk lines stay straight and rectilinear edge to edge, with subtle chromatic aberration near the corners. During the super-slow-motion beat, the same wide lens racks focus to a single giant butterfly hovering right at the lens, rendering it in crisp, detailed close-up while @fantasy-dragon and @woman soften slightly just beyond — no lens cut, a smooth focus pull. Immersive, body-near, action-cam-mounted-behind-the-rider intimacy with huge environmental width. ACTION TIMING 0.0s–3.0s — THE TALL PLUNGE & WATER SKIM: from directly behind, @fantasy-dragon tips over the very top of the immensely tall 90° waterfall and drops in a long near-vertical nose-down plunge, wings tucked tight and folded, streamlined and accelerating straight down the towering wall of falling water — flying down so close it almost grazes the rushing surface the whole way, spray and rainbow light streaking past. Near the forest floor it flares its wings wide open and arcs out of the dive, leveling inches above the glowing turquoise pool, wingtips and tail-fan slicing twin lines of spray, herons scattering ahead. @woman trails her fingers through the water, shrieking with joy. 3.0s–4.0s — INTO THE SWARM: @fantasy-dragon rises off the water with a few powerful wingbeats toward a vast cloud of giant luminous butterflies; as they approach, the swarm bursts apart and scatters in every direction — sapphire, amber and emerald wings whirling outward around them. @woman laughs, arms wide. 4.0s–6.0s — SUPER SLOW-MOTION BUTTERFLY (no cut): as they pass through the swarm, one large butterfly drifts up right to the lens and time ramps smoothly down into super slow motion, holding for about two seconds on a beautiful close-up of this single butterfly — its big iridescent wings beating slowly and gracefully, every detail visible: glowing sapphire-amber-emerald scales, fine vein patterns, light shining through the translucent membrane, tiny shimmering edges — while @fantasy-dragon and @woman glide on softly just behind it. Then, sharply and instantly, time snaps back to full real-time speed — a hard, sudden return to motion as the butterfly whips away and the flight surges forward again. 6.0s–9.0s — THE GLIDING WEAVE PAST THE MONKEYS: back at full speed, @fantasy-dragon spreads its wings wide and still and soars on locked wings, gliding silently between two massive crossing branches and a curtain of hanging lianas, banking and tilting on its spread wings — no flapping — leaves, fruit and flowers whipping past. Troops of long-limbed monkeys leap from tree to tree across the path and cling to the trunks, heads snapping round to stare after @fantasy-dragon and @woman in wide-eyed surprise as they glide by; parrots and hummingbirds scatter. @woman ducks low and lets out a delighted whoop. 9.0s–11.5s — THE NARROW GAP: still gliding, @fantasy-dragon folds its wings and shoots through a razor-narrow vertical gap between two leaning trunks — bark walls blurring past inches from the lens — then snaps its wings open on the far side in a corkscrew roll past a swooping flock. @woman throws her head back, laughing and cheering with pure exhilaration. 11.5s–15.0s — THE CLIMB AND REVEAL: @fantasy-dragon pitches up and powers straight upward along the colossal trunks with broad, heaving wingbeats, bark, branches, flowers and birds streaking downward past the lens as the camera (still behind) tilts up and cranes back. They burst up through the top of the canopy into open golden sky. The camera pulls back into a colossal wide establishing shot revealing the WHOLE world at once — an endless sea of mile-high trees rolling to the horizon, rivers of mist and the distant waterfall gorge between them, flocks like drifting confetti, @fantasy-dragon and @woman a bright speck rising above it all. @woman laughs breathlessly in wonder as @fantasy-dragon levels off and glances back at her with a soft chirp. Hold on the awe. PHYSICS @fantasy-dragon moves with the three wing modes above — wings tucked and streamlined in the long vertical plunge, wide and locked while gliding through the trees, broad and heaving while climbing. The flare out of the dive shows real aerodynamic load as the open wings catch the air and bend the fall into a level arc. In the slow-motion beat the butterfly's wingbeat reads in exquisite slow detail — wings flexing, bending and rippling through the stroke, scales catching light, body bobbing gently. Mass and inertia carry through every bank, roll, glide and level-out. Down the waterfall and over the water, wingtips and tail cut real spray trails and leave a wake; mist beads in the air. @woman's body shifts, ducks and braces with every move, weight transferring; her hair and clothes stream and snap with airspeed, gusts and spray. Branches, lianas, fruit, flowers, leaves and butterfly wings bend, scatter and recoil as they pass. Birds, butterflies, monkeys and ground animals react with real evasive and curious motion — monkeys leaping between trees and turning to look. Every motion carries weight and momentum with clear acceleration and deceleration. LIGHTING Primary light is warm golden sun high on screen-right, raking through the canopy in volumetric god-rays and catching mist, pollen, spores and butterfly wings. Maximum vibrancy and luminosity — deep emerald and jade greens, cobalt-to-turquoise sky and pool, warm gold accents, blazing pops of scarlet, magenta, violet, sapphire and coral from flowers, butterflies, birds and @fantasy-dragon. During the slow-motion close-up, sunlight glows through the butterfly's translucent wings, lighting up the iridescent color. At the waterfall, sun through the spray throws a vivid rainbow and bright speculars off the water. In the narrow gap the light goes briefly cool and shadowed, then bursts back into brilliant daylight on the far side. As they break above the canopy the light opens into bright open daylight across the whole landscape. Rich, saturated, glowing and painterly with deep, clean color. Sun rim-lights @fantasy-dragon's wings and @woman's hair; soft environmental bounce fills the shadows. AUDIO SFX only: the roar and hiss of the towering waterfall swelling louder as they plunge down it, a whoosh of rushing air in the long vertical dive, a soft thump of air as the wings flare open, spattering spray over the water, then soft fluttering as the swarm bursts apart. As time ramps into super slow motion the sound stretches low and dreamy — a deep, slowed flutter and a soft hush — then cuts back instantly with a sharp snap to real-time as full speed returns: a sudden surge of rushing wind, then the quiet rushing wind of the silent glide, chittering monkey calls as they leap and react, a sharp whoosh and echo through the narrow gap, broad steady wing-beats building on the final climb, @fantasy-dragon's soft friendly chirps (especially as it glances back), and @woman's delighted breath, laughter, gasps and joyful whoops throughout — wordless. Around them: rich birdsong, insect hum, rushing leaves and water. @fantasy-dragon's only vocal sound is its gentle chirp. The track stays fully natural and music-free from start to finish. POSITIVE LOCKS Colossal surreal scale held throughout — @fantasy-dragon and @woman are a tiny bright speck against mile-high trees and the towering waterfall, and the final wide reveals the entire landscape at once. The whole clip is one continuous take with no cuts: a smooth ramp down into the super-slow-motion butterfly close-up, then a sharp, instant snap back to full real-time speed. The camera stays behind them the whole time, holding their backs in frame, racking focus to the butterfly only for the slow beat. @fantasy-dragon is large with physically believable wings: tucked tight in the long 90° plunge almost grazing the water, opening near the ground, gliding wide and still between the trees, and beating broad and powerful only on the climb; it keeps the exact reference design every frame, and @woman is the same person throughout, matching the reference. Coherent, solid jungle and tree architecture with true perspective through the tall vertical waterfall, water skim, butterfly slow-motion, gliding weave and the narrow gap; vibrant glowing color and dense, varied wildlife — birds, butterflies, leaping curious monkeys, ground animals — all the way through; charming, physically believable flight with natural, film-like organic texture, lifelike surfaces, soft natural eyes, and a clean unmarked frame.

Look at the waterfall spray and the wing position as the creature touches the water. This is the iteration loop in action: generate, spot what you don't like, tell Claude exactly what to change, regenerate.看看生物触水时的瀑布水花和翅膀位置。这就是迭代循环:生成,发现不喜欢的地方,告诉 Claude 具体改哪里,重新生成。


Scene 3 - Glued to the screen场景 3 —— 盯着屏幕

Back to the room. All the assets already exist, so we just described the idea: the character is locked on the TV, mom calls him from off-screen, he mumbles "Yeah, one sec mom" and switches the channel.回到房间。所有素材都已经有了,所以我们直接描述想法:角色盯着电视,妈妈在画外喊他,他嘟囔着“这就来,妈”,然后换了台。

SCENE CONTEXT Evening, same living room. @ADIL sits on the sofa, fully fired-up by something on the TV we never see. From the unseen kitchen his mother calls him to come eat; too hyped to break away, he mumbles a delayed reply, grabs the remote off the low table, flips to another channel and keeps watching — he does not get up. ACTIVE REFERENCES @ADIL — young man, bright-eyed and hyped, immersed in the off-frame screen. 100% matches the reference. @REMOTE — ordinary household TV remote, resting on the low table in front of him at the start. 100% matches the reference. (Mother is an offscreen English voice only — no in-frame reference, never enters the shot.) LOCATION MAP @ADIL'S-ROOM — STYLE REFERENCE ONLY, not a fixed keyframe; model extends the world, never pinned 1:1. Foreground: the low coffee table with @REMOTE on it. Midground: the sofa, @ADIL facing forward toward the off-frame TV. Background: the room and a doorway/hallway leading off frame-right to the unseen kitchen; a side window frame-left as the fading light source. Lived-in room, faint haze in the air. FIRST FRAME / BLOCKING Non-empty opening frame: @ADIL already leaning forward off the sofa back, elbows on knees, eyes bright and locked on the off-frame screen, body bouncing slightly with excitement; @REMOTE sitting on the table in front of him. The off-frame TV sits directly ahead of the sofa, its glow on his face — never shown in frame. Asymmetric framing, subject already in motion. FORMAT MODE Timed multishot, explicit HARD CUTs at the stated seconds; cuts only at those points, the camera does not cut on its own. OPTICS Fixed shooting distance per shot — no zoom, no push-in; crop size changes only on the cut. SHOT 1 MS at 47°, SHOT 2 CU at 29°, SHOT 3 MS on his hands at 47°. No drift mid-segment. CAMERA Handheld throughout, live operator breathing, pronounced organic shake and micro-drift, a subtle dutch that resettles — never on sticks, no locked-off framing. Slight low angle in every shot, lens a touch below eye-level looking gently up at him. SHOT 1: slight low-angle handheld medium with restless sway riding his excited energy — feel his hyped momentum. SHOT 2: slight low-angle handheld close, jittery micro-drift holding on his grinning face through the delay — sit in the beat before he answers. SHOT 3: slight low-angle handheld on his hands as he grabs @REMOTE off the table and clicks, shaky drift as the glow shifts — he changes the channel and stays put. ACTION SHOT 1 (0:00–0:06) — @ADIL leans forward, eyes locked on the off-frame screen, a grin tugging, body bouncing slightly, screen glow dancing on his lit-up face. Offscreen his mother calls him to eat; he's too into it — no real reaction, maybe a tiny absent "mm". HARD CUT. SHOT 2 (0:06–0:11) — close on his face, still glued and grinning, light shifting across him; the offscreen voice calls again, sharper. A beat — his eyes flicker, he huffs a laugh, mumbles a reply without looking away. HARD CUT. SHOT 3 (0:11–0:15) — still locked on the screen, he reaches forward, picks @REMOTE up off the table, points it at the screen, thumb presses — soft click — the off-frame glow shifts hue as the channel changes; he settles back a touch, eyes never leaving the screen, still watching, not getting up. HARD CUT. PERFORMANCE Pure excitement and immersion, only half-acknowledging the call — genuine and human, never cartoonish. Bright engaged eyes with screen catch-lights, a grin tugging, small body bounce; on the second call a flicker in the eyes and a huffed laugh before the mumbled reply. Pore-level realism — vellus hair, capillary flush, visible breath and chest rise. PHYSICS @REMOTE carries real weight as he lifts it from the table. The cushion holds his lean. Correct contact shadows on the sofa and table. Nothing floats. LIGHTING Natural light only. Window key from frame-left, going dimmer and warmer toward evening, camera on the shadow side, faint atmospheric haze. The off-frame TV glow is the moving light on his face — cool, shifting hue on the channel change in SHOT 3. WB locked 5200K. COLOR GRADE Warm fading-daylight room neutrals as the dominant material, cool TV screen glow on his face as the secondary beam, with a chartreuse #d1ef17 accent living only in the @REMOTE standby LED, a flicker of the screen glow and a glint on a lens flare — color tied to source and surface, not a flat list. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. Room ambience, the soft compression of the cushion, a soft remote click, the low hum of the off-frame TV. Dialogue (English, audio only, no subtitles): Mother (offscreen): "Honey, dinner's ready! Come eat!" Mother (offscreen, sharper): "Did you hear me? Come on, it's getting cold!" @ADIL (mumbled, eyes still on screen): "Yeah— one sec, Mom!" STYLE 8K photorealistic — no 3D render, no game engine, no game-cutscene aesthetic. Naturalistic master cinematography, fine film grain. OUTPUT SETTINGS 16:9. Real-time throughout, no slow-motion. POSITIVE LOCKS Handheld shaky operator-feel in every shot, slight low angle throughout, fixed distance per shot with crop changing only on the cut — no zoom, no push-in. @ADIL stays seated, changes the channel and keeps watching — he does not stand up and does not turn the TV off. The off-frame TV is never in frame, not even blurred — felt only through the glow on his face. The mother stays offscreen, an English voice only. He picks @REMOTE up off the table before clicking; @REMOTE stays normal household remote size. Screen content stays unseen — no logos, no UI, no subtitles. Contact shadows on sofa and table read clearly. Wardrobe comes from the @ADIL reference and stays identical across all three cuts. Eyes stay natural — no eye glow.

Zoom in on the face: the pores, every single eyelash, the natural hair, even forehead wrinkles. It doesn't look AI-made - it looks like it was filmed on an expensive professional camera.放大看脸:毛孔、每一根睫毛、自然的头发,甚至额头的皱纹。这看起来不像 AI 做的——看起来像是用昂贵的专业摄像机拍出来的。


Scene 4 - The wildlife documentary场景 4 —— 野生动物纪录片

A super-detailed nature-documentary shot, like the ones on animal channels. First the location - a wide shot of snowy mountains - then the video: a snow leopard descending the slope, captured on a long lens from miles away.一个超详细的自然纪录片镜头,就像动物频道里那种。先是场景——雪山的广角镜头——然后是视频:一只雪豹走下山坡,从几英里外用长焦镜头拍摄。

4
Aerial cinematic photograph of a vast snow-covered mountain range, endless rows of sharp jagged peaks and ridges receding into hazy distance. Steep rocky slopes streaked with white snow and dark exposed stone, deep shadowed valleys between the ridges, thin wisps of cloud and blowing snow drifting around the summits. Cold muted blue-grey palette, soft natural daylight, atmospheric haze, high-altitude drone view looking across the peaks. Ultra-realistic, photorealistic, highly detailed, cinematic, no text, no logo.广袤雪山山脉的电影感航拍照片,无尽的锯齿状山峰和山脊退入朦胧的远方。陡峭的岩石斜坡上布满了白雪和裸露的深色岩石,山脊之间是深深的阴影谷地,薄薄的云团和飞雪在峰顶周围飘荡。冷淡的蓝灰色调,柔和的自然日光,大气雾霾,横跨山峰的高空无人机视角。超写实,照片级真实感,高细节,电影感,无文字,无 Logo。
SCENE CONTEXT A vast snowy mountain range in cold daylight. From one fixed vantage a super-telephoto lens zooms from a wide view of the peaks down to a snow leopard threading its way down the rocky ledges, settling on a half-body framing as breath steams from its nose. One continuous shot, 10 seconds. The leopard is on a distant slope across miles of air from the camera. ACTIVE REFERENCES @mountain — the location (source of truth): a vast range of steep snow-covered peaks, rocky ridgelines and ledges, layered summits fading into cold blue atmospheric haze, cold daylight. 100% matches the reference. THE SNOW LEOPARD (full description, identical every frame) A real snow leopard with thick pale grey-and-white fur dense with black rosettes and spots, pale-green eyes, a broad furry face, broad furry paws, and a very long thick spotted tail. Heavy plush winter coat, stocky muscular build. Keep its exact markings, colors and proportions consistent in every frame — no morphing, no drift. LOCATION MAP Foreground / midground: the steep rocky ledges and snowy steps the leopard descends. Background: layered snow-covered summits fading into cold blue haze. The camera is anchored far away on a single fixed vantage, looking across miles of distance toward the slope. FIRST FRAME / BLOCKING The camera opens at the wide 84° view, the mountains filling the frame and the snow leopard already a tiny speck moving on a distant slope — already descending, nothing frozen. Subject low and to one side of frame, the layered peaks reading top to bottom. FORMAT MODE One continuous shot (oner), 16:9, real-time, no cuts. The camera holds one fixed spot the whole time; only the lens focal length changes — pure optical zoom, the camera does not cut on its own. OPTICS A single massive optical zoom from WIDE 84° FOV to super-telephoto 8° FOV, stopping at a HALF-BODY framing of the leopard (head, chest and shoulders, roughly half the body). Extreme compression at the long end — mountain layers flattened into stacked planes. Genuine long-distance super-tele character: from the moment the zoom pushes long, and strongly by the close end, the image swims and ripples with heat-haze and atmospheric distortion, edges shimmering as if seen through rising air over miles; a constant nervous telephoto micro-tremor, the framing trembling and drifting; the picture grows softer and slightly mushy, fine sharpness falling off into gentle blur and faint digital-zoom smear with light grain; a brief focus hunt before it settles on the cat. The wide start is clean and crisp; these artifacts build hard as it zooms in. Continuous push, no drift between beats. CAMERA Fully static locked-off tripod, anchored on the shadow-and-distance side of the slope, operator axis fixed across the valley. The camera never repositions, dollies or pans — it stays in ONE spot; the only movement is the lens zooming long. 180° shutter motion blur. Wide tonal latitude holding both bright snow highlights and shadowed rock without clipping, gentle highlight roll-off on the lit snow. ACTION Throughout, the snow leopard descends the mountain — picking its way down the rocky ledges and snowy steps at a slow, careful walking pace, sure-footed and deliberate, body low, the long tail counterbalancing, placing each broad paw ledge to ledge. Small clumps and flecks of snow cling to its coat, some shaking loose as it moves. Camera and subject motion stated separately on the 10-second clock: 0.0s–2.5s — camera at the wide 84°, the leopard a tiny speck descending a distant slope. 2.5s–7.5s — a continuous optical zoom-in from 84° toward 8° (camera staying put), the heat-shimmer, telephoto jitter and softness all intensifying as the lens pushes long; the leopard keeps descending. 7.5s–10.0s — the zoom settles and holds at a HALF-BODY framing, the image trembling, soft and shimmering; the leopard pauses, scanning, warm breath vapor puffing and steaming from its nose and mouth with each exhale, curling away in the cold. PERFORMANCE Real big-cat behavior — careful, deliberate downward steps, body low and sure-footed, long tail swaying to counterbalance, broad paws set ledge to ledge, pausing to scan with breath fogging, fully alive from frame one. Pore-level fur realism: thick pale grey-and-white coat dense with black rosettes, snow flecks caught in the fur of shoulders, legs and paws, warm breath steaming from nose and mouth, living pale-green eyes with cold catch-lights. PHYSICS Gravity and inertia respected — balanced downward steps, paws gripping rock and snow, loose snow sliding underfoot, the long tail swaying for balance. Warm breath condenses into visible vapor in the cold air. Snow flecks on the fur jostle and drop with movement. LIGHTING Cold natural daylight, white balance locked at 5600K daylight with a cold cast. Blue atmospheric haze layering the peaks, soft snow highlights, the leopard's pale fur blending into rock and snow, warm breath vapor catching the cold light. Crisp at the wide start; hazier and softer when zoomed long. Atmospheric haze builds across the shot — roughly 20% density at the wide start rising to about 70% at the long compressed end, visible deep into the layered peaks. COLOR GRADE 60:30:10 — dominant cold snow-and-rock neutrals; secondary cold blue atmospheric haze blanketing the layered peaks; accent the warm steam of the leopard's breath catching the light and its pale-green eyes. Cold neutrals as material under flat daylight, the blue as the air itself thickening with distance, the warm breath as the single live spark of heat. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. High-altitude wind, the faint slide of snow and loose rock under the paws, soft audible breathing at the close framing, distant mountain ambience, the muffled quiet of altitude. STYLE 8K, photorealistic, live-action long-distance super-telephoto wildlife capture — real footage, not a render or game engine. Naturalistic cinematography, one continuous locked-off zoom, fine natural grain building with the long lens. OUTPUT SETTINGS 8K, 16:9, 24fps, real-time throughout, total duration 10 seconds, one continuous shot. POSITIVE LOCKS The camera stays in one fixed spot — static tripod, pure optical zoom only, never cuts and never leaves its spot. The snow leopard keeps the exact same fur, markings, pale-green eyes and proportions throughout — no identity drift; the location stays exactly @mountain, same mountain range throughout. One continuous optical zoom from 84° wide to 8° super-tele, stopping at a HALF-BODY framing. The long-lens look stays clearly realistic and imperfect — visible heat-haze swim, telephoto jitter, soft mushy grain, all building as it zooms in. The leopard descends the rocky ledges sure-footed, with visible breath vapor and snow flecks in its fur. 10 seconds, real-time.

Look at the fur, the paws, the ears, the breath fogging in the cold air. Zoom in on the leopard - instead of pixels you see fur, like a macro shot.看看皮毛、爪子、耳朵,还有冷空气中呼出的白气。放大看雪豹——你看到的不是像素,而是皮毛,就像微距摄影一样。

Then we made it harder: the shot starts super wide with the leopard barely visible, and the lens slowly zooms all the way in to a half-body close-up. One more tip: since this is a documentary, add a voiceover. We opened Supercomp, uploaded the video, and asked it to analyze the clip and write a wildlife-documentary voiceover - then picked a voice. No script-writing stress.然后我们增加了难度:镜头从超广角开始,雪豹几乎看不见,然后镜头缓慢变焦,一直拉近到半身特写。 再给个建议:既然是纪录片,加个旁白。我们打开 Supercomp,上传视频,让它分析片段并写一段野生动物纪录片风格的旁白——然后选个声音。不用为写稿发愁。

Analyze the video and create a voiceover for it in the style of wildlife documentaries分析视频,并以野生动物纪录片的风格为它创作旁白

Scene 5 - The truck vs the snow monster场景 5 —— 卡车对决雪怪

The big one: an epic wide shot of a truck outrunning a snowstorm with a monster inside it. Assets needed: the snow monster and the truck. 重头戏:卡车在暴风雪中狂奔,车里藏着怪物的史诗级广角镜头。需要的素材:雪怪和卡车。

5
CHARACTER REFERENCE SHEET PROMPT — SNOW/ICE MONSTER (3 views) Production character turnaround / reference sheet of a single creature, presented as three clean shots side by side on a neutral seamless grey background, even soft studio lighting, sharp focus, minimal atmosphere, no environment, no truck — concept-art reference style. The same creature in all three panels, perfectly consistent design. Panels labeled: PORTRAIT (left), FULL BODY — FRONT (center), FULL BODY — REAR (right). The creature: a towering humanoid blizzard monster formed from compacted, churning snow and dense ice — a colossal frostbitten giant. Its body is sculpted from layered packed snow and translucent blue-white ice, surfaces crusted with frost crystals, jagged icicle growths, and wind-carved ridges, with faint wisps of loose snow drifting off its silhouette so its edges look half-dissolved and storm-born. Massive broad shoulders, long heavy arms ending in oversized clawed hands of fused ice shards, powerful legs and large crushing feet of compacted snow and ice. Pale, softly glowing almond-shaped eyes set deep in a frost-rimed face, a wide maw lined with rows of jagged translucent icicle teeth. Cold blue-grey, steel-blue, and white palette, internal pale glow, hard cold speculars on the ice. Portrait (left): head and shoulders close-up — frost-crystal face, glowing pale eyes, icicle teeth, textured ice surfaces, menacing. Full body front (center): full standing figure head-to-feet, facing camera, arms and clawed hands and legs and feet all visible, symmetrical, showing full scale and proportions. Full body rear (right): full standing figure from behind, head-to-feet, showing the back, shoulders, spine ridges of jagged ice, arms, and legs. 8K photorealistic, hyperdetailed, fine grain, tactile ice and snow texture, true concept-art clarity, consistent design and scale across all three panels.角色参考表提示词 — 雪/冰怪(3 个视角) 单一生物的制作角色转面图/参考表,在无缝中性灰背景上并排展示三个清晰镜头,均匀柔和的摄影棚光,锐利对焦,最小化环境,无环境,无卡车——概念艺术参考风格。三个面板中的生物相同,设计完美一致。面板标签:肖像(左)、全身正面(中)、全身背面(右)。 生物:由压实的搅动雪和致密冰组成的耸立人形暴风雪怪物——巨大的冻伤巨人。身体由层层压实的雪和半透明蓝白色冰雕刻而成,表面覆盖着霜晶、锯齿状冰柱生长物,风雕刻的脊,身上飘散着微弱的松散雪花,边缘看起来半溶解,仿佛诞生于风暴。巨大的宽肩膀,长而沉重的手臂,末端是巨大的冰碎片融合而成的爪子,强壮的腿和由压实的雪和冰组成的巨大粉碎性脚。苍白、柔和发光的杏仁眼深陷在结霜的脸庞中,宽大的嘴巴里排列着锯齿状半透明冰柱牙齿。冷蓝灰色、钢蓝色和白色色调,内部苍白发光,冰上有坚硬的冷高光。 肖像(左):头部和肩膀特写——霜晶脸,发光的苍白眼睛,冰柱牙齿,纹理冰面,充满威胁感。 全身正面(中):完整的站立人物,从头到脚,面向镜头,手臂、爪子、腿和脚都可见,对称,展示完整的比例和体型。 全身背面(右):完整的站立人物,从背后看,从头到脚,展示背部、肩膀、锯齿状冰脊、手臂和腿。 8K 照片级真实感,超细节,细颗粒,触感冰雪纹理,真正的概念艺术清晰度,三个面板设计和比例一致。
5
PROP REFERENCE SHEET PROMPT — SNOW-TRUCK (4 views) Production prop turnaround / reference sheet of a single vehicle, presented as a 2x2 grid of four clean studio shots on a neutral seamless grey background, even soft product lighting, sharp focus, no atmosphere, no snow, no characters — pure orthographic reference style. The four panels, labeled: FRONT VIEW (top-left), SIDE VIEW (top-right), REAR VIEW (bottom-left), BOTTOM / UNDERCARRIAGE VIEW (bottom-right). The same vehicle in all four panels, perfectly consistent design. The vehicle: a heavy armored post-apocalyptic Mad-Max-style snow-truck built for deep snow. Boxy welded steel cab with riveted armor plating, scavenged battered panels, a roof rack of round amber/orange running lamps and a tall whip antenna, twin side mirrors, and a heavy steel snowplow blade angled across the nose. Big greenhouse windshield. Oversized deep-lug studded snow tires up front with aggressive blocky tread, massive studded rubber snow-tracks at the rear for flotation, exhaust stacks venting from behind the cab, a covered cargo box behind the cab. Weathered, ice-crusted, lived-in industrial scrap aesthetic — steel-grey, gunmetal, rust, with cold blue-grey tones. Front view: straight-on head-on, snowplow blade, front tires, lamp rack, windshield, symmetrical. Side view: full broadside profile showing cab, cargo box, front tires and rear snow-tracks, overall proportions. Rear view: straight-on back, cargo box doors, taillights, rear snow-tracks, exhaust. Bottom view: full underside/undercarriage looking up — chassis frame, axles, drivetrain, suspension, track mechanism, fuel tank, skid plates. 8K photorealistic, hyperdetailed, fine grain, true product/concept-art clarity, consistent scale and design across all four panels.道具参考表提示词 — 雪地卡车(4 个视角) 单一车辆的制作道具转面图/参考表,在无缝中性灰背景上以 2x2 网格展示四个清晰的摄影棚镜头,均匀柔和的产品光,锐利对焦,无环境,无雪,无角色——纯正交参考风格。四个面板标签:正面(左上)、侧面(右上)、背面(左下)、底部/底盘(右下)。四个面板中的车辆相同,设计完美一致。 车辆:重型装甲末日废土风雪地卡车,专为深雪打造。箱型焊接钢驾驶室,铆接装甲板,车顶架上有圆形的琥珀色/橙色行车灯和高鞭状天线,双侧后视镜,车头斜挂着重型钢制扫雪铲。巨大的温室挡风玻璃。前轮是超大深花纹雪地轮胎,后轮是巨大的带钉橡胶雪地履带,排气管从驾驶室后方伸出,驾驶室后方有一个封闭的货箱。风化、结冰、有生活气息的工业废料美学——钢灰色、枪灰色、锈迹,带有冷蓝灰色调。 正面:正前方,扫雪铲,前轮胎,灯架,挡风玻璃,对称。 侧面:完整的侧面轮廓,显示驾驶室、货箱、前轮胎和后雪地履带,整体比例。 背面:正后方,货箱门,尾灯,后雪地履带,排气管。 底部:完整的底部/底盘仰视——底盘框架、车轴、传动系统、悬挂、履带机构、油箱、防滑板。 8K 照片级真实感,超细节,细颗粒,真正的产品/概念艺术清晰度,四个面板比例和设计一致。
SCENE CONTEXT In a frozen post-apocalyptic wasteland, @truck1 — an armored snow-truck — races down a white highway at full power, weaving through a frozen graveyard of wrecks and ice barricades while desperately trying to outrun a massive blizzard-avalanche that chases it from directly behind. The man is the driver, both hands on the steering wheel; the woman sits beside him in the passenger seat and does not touch the wheel. @monst, a colossal ice monster, emerges out of the churning blizzard behind the truck, dives beneath the snow like a shark and surfaces still behind it, lunges and just misses with its claws, then heaves a massive hand under the truck and flips it. The truck tumbles, lands and rolls toward a low ground-level camera, coming to rest upside down right in front of the lens. Grounded, photoreal, high-budget fantasy realism — every surface, particle, and performance reads as captured live, not generated. ACTIVE REFERENCES @truck1: the heavy armored Mad-Max-style snow-truck — welded steel cab with riveted armor plating, roof rack of amber running lamps and whip antenna, steel snowplow blade at the nose, oversized deep-lug studded snow tires up front, studded rubber snow-tracks at the rear, exhaust stacks, covered cargo box. Weathered, ice-crusted, steel-grey. 100% matches the reference. @monst: the towering humanoid blizzard monster of compacted churning snow and translucent ice — frost-crystal surfaces, jagged icicle growths, pale softly glowing almond eyes, a maw of icicle teeth, massive clawed hands of fused ice shards. Edges half-dissolved into drifting snow, never fully solid; it forms from and emerges out of the blizzard, can submerge beneath the snow and resurface, and relentlessly chases the truck from behind. 100% matches the reference. Cab occupants: the man is the driver in the screen-left seat, both hands gripping the steering wheel, actively driving. The woman is the passenger in the screen-right seat; she only sits in the cab and never holds or reaches for the wheel — her hands stay on her lap, the dash, or the door. Two survivors in scavenged cold-weather gear. LOCATION MAP Wide frozen plateau. A snow-packed highway runs horizontally across the frame, entering from the deep-background left edge (x 5%, y 42%) and sweeping toward the right edge, so all travel reads clearly left-to-right; the road ahead (screen-right) is open and clear. Along the entire length of the road, from the far establishing distance onward: a scattered frozen convoy of abandoned, ice-encased trucks and cars — half-buried in drifted snow, some overturned on their roofs or sides, rusted and frost-covered, the wreckage of those who came before. Plus leaning ice ridges and packed-snow barricades, a collapsed highway sign sagging across the lane, dead ice-crusted utility poles with sagging wires. The 200-meter-tall wall of churning white blizzard/avalanche is always behind the truck — anchored to screen-left and the rear horizon, advancing rightward in pursuit but never ahead of the truck and never in its direction of travel. @monst lives inside this rear storm wall and rises out of it, always trailing behind the truck. Weak sun glow hidden behind the storm wall acts as backlight. Lighting direction: diffuse top-down with cold backlit haze from behind/left. FIRST FRAME AND SPATIAL BLOCKING The first visible frame is an extreme super-wide aerial-height establishing vista — maximum scope, @truck1 a tiny speck. The truck sits deep background, left-of-center, x 38%, y 44%, scale tiny (under 4% of frame width), driving left-to-right at full speed. Already visible in this wide establishing frame, scattered along the roadside and the snow plain, are abandoned snow-covered trucks and cars — some upright and half-buried, some overturned — dotting the route ahead and behind, establishing the frozen graveyard the truck will weave through. The blizzard wall fills the background behind and to the left, x 0% to 70%, upper half of frame, advancing rightward — always behind the truck, never in the open road ahead. No empty plate — the truck is present and moving left-to-right in frame one. The camera pushes in and tracks right with the truck until the cab and windshield dominate, x 50%, y 50%, the man driving (screen-left seat) and the woman passenger (screen-right seat) clearly visible through the glass. Later @monst surges out of the storm behind the truck. Truck travel direction: screen-left to screen-right, locked; clear road ahead at screen-right. Storm and monster locked to screen-left/rear (behind). Male driver: torso forward (screen-right), both hands gripping the steering wheel, eyes flicking to the side mirror back toward the storm (screen-left). Female passenger: torso forward, head snapping back toward @monst (screen-left) then front; hands brace on the dash or lap, never on the wheel. Both bodies face direction of travel (screen-right). FORMAT MODE Controlled six-segment sequence in one continuous escalating beat. Seg 1 real-time push-in. Seg 2 real-time obstacle gauntlet. Seg 3 real-time "shark-under-snow" stalk. Seg 4 slow-motion fake-out (claw miss). Seg 5 slow-motion scoop, launch, and mid-air rotations. Seg 6 low ground-level shot as the truck rolls in and stops upside down. HARD CUT at 3.0s, HARD CUT at 6.0s, SMASH CUT to slow motion at 9.0s, continue slow motion through the scoop, MATCH CUT to the ground angle at 13.5s. No subtitles, no music. OPTICS Prime-lens optical character throughout — clean, sharp center field, gentle natural falloff, no zoom-breathing, fixed-focal clarity. LENS LOCK SEG 1 = begin 107° wide rectilinear (immense environment, straight horizon, truck tiny, no fisheye), settling toward 84° classic wide as it closes on the cab. No drift beyond this push. LENS LOCK SEG 2 = 84° classic wide, low and close, foreground wrecks and barricades looming and ripping past the lens, immersive speed, straight lines rectilinear. LENS LOCK SEG 3 = 47° standard normal side profile, truck and the moving snow-bulge both readable, grounded perspective. LENS LOCK SEG 4 = 29° short telephoto, slight compression on the near-miss claw and the heroes' faces, the monster looming soft behind. LENS LOCK SEG 5 = 47° standard normal for the launch/flip — natural human-eye proportions, no distortion as the truck tumbles. LENS LOCK SEG 6 = 84° classic wide, low ground-level — the truck looms large rolling toward the lens, foreground snow road across the lower frame, no fisheye. CAMERA ARRI Alexa large-format cinema-camera character — wide tonal latitude, gentle highlight roll-off in the bright snow, clean shadow detail, true filmic color science. Seg 1: high aerial height over the plateau for max scope, then descends and pushes in while tracking right with @truck1, accelerating to match it, ending at windshield height ~2 m off the front-left quarter. Powerful weighted aerial-to-chase glide, subtle operator weight, no jitter, no drone float. Seg 2: HARD CUT to a low fast tracking shot riding just ahead of and beside the plow blade as the truck smashes and weaves through the wreck field — foreground debris whipping past, a brief duck under the collapsed sign. Seg 3: HARD CUT to a side tracking shot pacing the truck's flank; camera tilts back toward the rear to catch a fast snow-bulge wave coursing under the road surface behind the truck, then whips up as @monst bursts out of the snow behind it. Seg 4: SMASH CUT to slow motion — a tighter three-quarter rear angle on the cab as the monster's clawed hand lunges in from behind and rakes past, just missing; camera catches the heroes' faces. Seg 5: stay in slow motion, ease wider as the hand plunges into the ground behind/under the truck and heaves it up; camera holds the airborne truck and follows it through its rotations, debris and snow crossing the lens. Seg 6: MATCH CUT to a static low-angle ground shot — camera planted on the snow road, lens just above the surface, locked off. The truck drops in, lands, and rolls straight toward camera, growing huge, then settles upside down right in front of the lens. No camera move. ACTION TIMING 0.0s to 3.0s — Seg 1: Extreme super-wide establishing vista, tiny @truck1 flooring it left-to-right along a road dotted with abandoned, snow-covered and overturned trucks and cars, blizzard wall surging from screen-left behind. Continuous push-in and right-track resolves to a close-up of the two heroes — the man white-knuckle on the wheel, the woman glancing back. @monst's face surfaces from inside the churning wall behind, gaining. 3.0s HARD CUT. 3.0s to 6.0s — Seg 2: Obstacle gauntlet. The steel plow blade smashes through a packed-snow barricade in an explosion of powder, the truck swerves hard around the ice-encased hulk of a dead semi, clips a torus of ice debris, and ducks low under a collapsed highway sign. Snow spray, ice shards, real speed and near-misses. 6.0s HARD CUT. 6.0s to 9.0s — Seg 3: The shark stalk. @monst dives beneath the snow behind the truck and disappears; a long, fast-moving bulge-wave of churning snow races under the road surface behind the truck, gaining — the heroes spot it in the mirror, dread rising. The wave surges, then @monst erupts upward out of the snow in a tower of powder, hauling its towering body free of the storm directly behind @truck1. 9.0s SMASH CUT to slow motion. 9.0s to 11.0s — Seg 4: The fake-out. In slow motion @monst lunges from behind and swings its huge clawed hand at the truck — the claws rake through the air and just miss the cargo box by inches, tearing only snow and a side mirror. For a held beat the heroes' faces flood with split-second relief, believing they've broken free. 11.0s to 13.5s — Seg 5: The reversal. Still in slow motion, the monster's hand drops and plunges into the ground behind and under the truck, scooping up a wave of frozen earth and snow that levers @truck1 into the air. The truck tumbles through two full mid-air rotations, glass and snow spraying outward, the heroes' relief snapping back to terror. 13.5s MATCH CUT to the low ground-level angle. 13.5s to 16.0s — Seg 6: Static low ground-level shot. The truck drops into frame, hits the snow road and rolls toward camera in a spray of powder, decelerating, and comes to rest upside down right in front of the lens, filling the frame. One front wheel keeps spinning freely for a beat, then slows and stops. Loose snow drifts down, steam and dust settle. PERFORMANCE (Hollywood-grade acting) Naturalistic, restrained, high-stakes performances captured like seasoned A-list film actors, with visible tension and real fear riding under the control. The male driver: jaw clenched, brow tight, a hard swallow, eyes wide and darting between the road, the mirror, and the swerving wrecks, white-knuckle grip as he muscles the wheel — fear held back by focus. The woman passenger: breath catching, eyes wide tracking the snow-bulge and the monster behind, a flinch, a hand braced hard on the dash; she never touches the wheel. On the fake-out, both faces flash a half-second of desperate relief — a gasp, the smallest easing — instantly ripped back into raw terror as the truck is launched, heads and limbs thrown against belts. Genuine adrenaline and dread throughout, no stiff or doll-like faces, no exaggerated mugging. Two survivors who have done this before, scared but capable. PHYSICS Driving on snow: @truck1 rides on purpose-built winter rubber — massive deep-lug snow tires with aggressive blocky tread, sipes and metal studs biting into the packed snow, plus studded rear snow-tracks for flotation. Tires visibly compress and grip, deforming against the snow, finding traction and occasionally breaking loose into brief controlled slip; tread packs and sheds snow each rotation. The heavy chassis sits low and planted, suspension compressing over ridges and lurching through swerves, throwing continuous white snow-spray and powder rooster-tails. Exhaust stacks vent hot steam shearing backward (screen-left). Engine at full strain, genuinely heavy mass and momentum. Obstacles: the plow blade meets each barricade with real impact — packed snow shatters and sprays, ice slabs crack and tumble with weight, the truck's mass barely deflects as it punches through; swerves load the suspension and shift the chassis with believable inertia and grip-loss. Shark stalk: @monst submerged under the snow displaces a raised, fast-moving bulge of churning powder that plows forward like a wake behind the truck, snow cascading off its crest; on the burst-out, a column of snow erupts upward with real mass and gravity as the monster heaves free, lower body still trailing back into the churning storm. Fake-out: the clawed hand carries real mass and speed as it swipes through the air just behind the truck, shredding a curtain of snow and clipping the side mirror, missing the body — believable near-contact, wind and powder blasting off the truck in its wake. Launch and crash: the scooping hand levers @truck1 upward by an erupting wave of frozen earth and snow. It leaves the ground carrying its full mass and load — heavy engine block, armor plating, fuel weight pulling the center of gravity low and forward — so it rotates with believable, slightly sluggish angular momentum, not a light flip. During the airborne rotations every wheel and track stays firmly mounted in its correct factory position — wheels keep spinning under their own inertia but never detach, dislodge, shift along the axle, droop, or float out of place; front tires stay on the front axle, rear tracks stay at the rear, all geometry intact. Panels and glass deform and shatter in slow arcs, loose gear and debris fly with correct mass and gravity. On landing, the heaviest mass (engine/front) makes first hard contact, the roof and cab crumple under the full weight as the chassis slams down and rolls toward camera with real momentum, decelerating against the snow's friction, settling upside down with real weight and load distribution. The wheels remain mounted; one front wheel keeps spinning freely on its axle, then slows under friction and stops. No floaty, weightless, or rubbery motion. Storm: blizzard wall churns with turbulent rolling motion, particles driven forward by wind, denser at the base, always closing on the truck from behind. @monst: a towering form of compacted swirling snow and ice — jagged ice teeth, hollow pale-glowing eye-sockets, frost-crystal features, a massive hand of fused ice shards that plunges into the ground and heaves it upward. It coalesces and partly dissolves, emerging from the rear storm and never fully solid, moving with the storm; its scoop carries crushing earth-moving force. Bodies: the driver's arms carry the steering wheel's weight; the woman's hair and scarf lag and whip; through the launch and flips both bodies obey real inertia. COLOR GRADE AND ATMOSPHERE Match the reference frame: a cold teal-and-cyan cinematic grade, heavily desaturated, low saturation across the board with everything pushed toward steel-blue and blue-green. Crushed, slightly lifted blacks holding milky shadow detail in the haze, soft filmic highlight roll-off, gentle low contrast from dense atmosphere. Thick volumetric blizzard haze and drifting snow particles fill the depth, softening every edge and stacking misty layers between camera and background so @monst reads as a ghostly, half-dissolved silhouette of textured ice rising out of the rear storm, with two pale, softly glowing almond eyes and rows of translucent icicle teeth looming over the scene. Light wraps and diffuses through the snow-fog. The truck's small practical lamps — amber/orange running lights and warm headlights — are the only warm accents in the frame, glowing as muted pops against the overwhelming cold blue palette, slightly bloomed through the haze. Overall mood: ominous, monumental, frostbitten, painterly cold — Roger Deakins-style atmospheric teal twilight. LIGHTING Cold overcast daylight, flat and diffuse, exposed for the bright snow field with Alexa-style highlight roll-off so the snow holds detail. Hidden sun glow behind the rear storm wall backlights the blizzard so @monst reads as a brighter churning silhouette against the gloom. Truck surfaces catch cool top light with hard cold speculars on wet metal and ice; cab interior falls into cooler shadow, faces revealed by soft windshield bounce, dash glow, and the spill of the warm practical lamps. Cold blue-cyan key everywhere; the only warm light is the truck's amber running lights and headlights as small contained accents — no broad warm key, no studio fill. WARDROBE (Mad Max, snow-logical) Male driver: scavenged cold-weather warlord gear — fur-lined battered leather greatcoat over a quilted thermal parka, riveted metal pauldron on one shoulder, fingerless gloves over thermal liners, frost-fogged ski goggles pushed up on a fur-flap ushanka cap, respirator scarf at the neck, bone and scrap-metal ornaments. Weathered, ice-crusted, lived-in. Female passenger: hooded fur-collared shearling jacket cinched with leather harness straps and buckles, knitted balaclava pulled down, goggles around the neck, fingerless gloves, scavenged-metal arm guards. Practical, insulated, battle-worn — same scrap-and-fur aesthetic, never impractical for sub-zero cold. STYLE Large-scale fantasy-realism in the vein of George Miller's kinetic wasteland chase staging, Denis Villeneuve's monumental scale and cold restraint, Peter Jackson's tactile creature physicality, and Zack Snyder's heroic backlit silhouettes and slow-motion impact beats. Photoreal texture, real grain, true film color science, grounded physical cinema — epic scope that still feels captured on a real set. 8K photorealistic, anamorphic widescreen, epic dark-fantasy cold blockbuster grade (Peter Jackson / Villeneuve weight), fine grain, hyperdetailed. OUTPUT SETTINGS IMAX-grade 8K capture, high bitrate 80–120 Mbps, ARRI Alexa large-format sensor character, prime-lens rendering, anamorphic widescreen, real-time motion in Segments 1–3, slow motion in Segments 4–5, ground-level static in Segment 6. POSITIVE LOCKS On the opening super-wide establishing shot, abandoned snow-covered trucks and cars (some overturned) are already visible scattered along the road and plain, making the later wreck-gauntlet logical and consistent. The man is the driver in the screen-left seat with both hands on the steering wheel, driving the whole shot. The woman is the passenger in the screen-right seat; she only sits and never holds, grabs, or reaches for the steering wheel. The snow blizzard/avalanche is ALWAYS behind the truck, chasing it — it never appears ahead of the truck or in its direction of travel; the road ahead (screen-right) stays clear. @monst is always behind the truck and always appears out of the storm — it emerges from the blizzard, submerges beneath the snow like a shark and travels as a moving snow-bulge wave behind the truck before bursting out, never a separate creature on clear open ground and never ahead of the truck. The chase passes through a real obstacle gauntlet — the plow blade smashes through ice barricades, the truck weaves around a frozen convoy of wrecks and ducks under a collapsed sign. Before the finale there is a fake-out: the monster's clawed hand lunges from behind and just misses the truck, the heroes feel a half-second of relief, then the hand digs under the truck and heaves the ground upward, launching it. The monster does not destroy the truck with claws; the flip comes from the under-scoop. The launch and two mid-air rotations play in slow motion with the truck's real mass and load — heavy, weighted, never floaty. During the entire airborne flip the wheels and tracks stay fixed in their correct mounted positions, spinning but never detaching or shifting. The final beat is a static low-angle ground shot with the camera on the snow road: the truck rolls in toward the lens and stops upside down right in front of camera, one front wheel still spinning before it slows and stops. @truck1 runs on special deep-lug studded snow tires and snow-tracks, travels left-to-right, and the shot opens on an extreme super-wide establishing vista before pushing in to the heroes. Cold teal/cyan desaturated grade with thick blizzard haze; @monst reads as a ghostly ice silhouette with pale glowing eyes; the truck's amber lamps are the only warm accents. Faces carry visible tension and fear; performances stay photoreal and restrained, no stiff or artificial faces.

Just watching it makes you feel like you are standing right there in the middle of the blizzard. It’s all because of the impressive wide shot.光是看着它,就让你感觉自己正站在暴风雪的中心。这全靠那个令人印象深刻的广角镜头。

The hardest part of this entire shot is the exact moment the truck flips over. Normally, in standard 1080p, this is where the whole generation would just glitch out, but here, the dynamics are insane. Look at the impact physics when the truck smashes through everything in its path. The slow-mo, the snow spray, and the way the debris actually fade into the distance instead of just melting.整个镜头最难的部分是卡车翻滚的那一刻。通常在标准的 1080p 下,这地方生成就会直接崩掉,但在这里,动态效果简直疯狂。看看卡车撞碎路径上一切时的冲击物理效果。慢动作、雪花飞溅,还有碎片真正消散在远方而不是直接融化消失。

This looks like a massive post-apocalyptic action film, and all it took was just generating a monster, a truck, and one video prompt.这看起来像是一部宏大的末日动作片,而这一切只需要生成一个怪物、一辆卡车和一个视频提示词。

Okay, Scene 5 is cooked up, and it's time to show you one more pro tip. Let's place this snow scene inside the TV that our main character is watching. To do this, we'll split the shot and take the first 6 seconds.好了,场景 5 搞定了,是时候告诉你另一个专业技巧了。让我们把这个雪景放到主角正在看的电视里。 为此,我们要分割镜头,取前 6 秒。

That's going to be our video reference that plays on the TV.这将是我们电视里播放的视频参考。

Before, you'd need VFX to pull off an effect like this - but we'll do it all with AI.以前,你需要 VFX 才能实现这种效果——但现在我们全用 AI 完成。

SCENE CONTEXT A lived-in living room toward evening. A man sits on the front edge of the sofa facing a flatscreen TV across the room, leaning forward toward the screen with a remote loosely in hand, watching a video playing on the TV. The camera is anchored behind and slightly to one side of the sofa, looking over his lowered head toward the fully visible screen. One continuous static 6-second shot. ACTIVE REFERENCES @ADIL — the man, seen from behind and three-quarter back: leaning slightly forward, perched on the front edge of the sofa, off the backrest, back and shoulders and back of head to camera, low and to one side so he never blocks the TV. Same wardrobe throughout. 100% matches the reference. @REMOTE — the remote: normal household size, held loosely in his hand, pointed loosely toward the TV but not pressed. 100% matches the reference. @ADIL'S-ROOM — STYLE REFERENCE ONLY, not a fixed keyframe: the same lived-in living room, sofa facing the flatscreen across the room, soft daylight fading through the side window, faint haze in the air. Extend the world naturally; do not reproduce the reference frame 1:1. @video_1 — the footage playing on the TV: on-screen content matches the source 1:1 — same image, timing, framing, motion and colors, never reinterpreted, re-edited, cropped, looped early, or replaced. LOCATION MAP Foreground (low, frame-left): the back of the man's head and shoulders, out of focus, rim-lit by the TV; @REMOTE the remote in his hand. Midground: the sofa front edge and the floor between sofa and TV. Background (center, sharp): the flatscreen on the far wall showing @video_1, keeping the exact same screen size and position it has in the @ADIL'S-ROOM room; the side window frame-left spilling fading daylight. Camera anchored behind the sofa on the shadow side. Warm room neutrals dominate, cool TV-screen glow as the secondary tone, a chartreuse #d1ef17 accent on the remote standby LED and a glint of screen glow. FIRST FRAME / BLOCKING From frame one the TV is already playing @video_1 and is fully visible. The man is already seated on the front edge, leaning forward toward the screen, remote in hand, head low in the foreground, breathing and faintly shifting — nothing is frozen. Rule of thirds: head and shoulders sit low and to one side of the foreground so the flatscreen stays fully visible and unobstructed center-frame. FORMAT MODE One continuous shot, the camera does not cut and does not move on its own. OPTICS Single static medium-wide at 47° FOV, framing the man's head and shoulders in the lower foreground and the whole flatscreen unobstructed in the background. Real cine-lens character, 180° shutter motion blur. No drift. CAMERA Locked off behind the sofa on the shadow side, eye-level-low, slightly off the subject's axis so his head sits low and to one side and never overlaps the TV, and so the screen is seen at a slight off-axis angle. Focus held on the TV screen, the foreground head soft. No pan, no push, no shake. Wide tonal latitude, soft highlight roll-off holding the bright screen. ACTION 0.0s to 6.0s — @video_1 plays on the TV continuously and unchanged, matching the source 1:1. The man sits leaning forward toward the screen, perched on the front edge of the sofa, off the backrest, the remote @REMOTE resting in his hand, watching. He breathes, his shoulders settle and shift a little, the back of his head and hair catch the cool shifting glow of the screen. The camera stays fully static the entire time. PERFORMANCE Top-tier cinematic restraint — small live shifts of the shoulders and head, slow natural breathing, a deliberate human stillness while watching. Pore-level realism: vellus hair, asymmetric moles, capillary flush, pore-shadows matching the window and screen light, rim glow on the hair from the TV. PHYSICS Real body weight on the front edge of the cushions, the cushions compressing where he sits, clear contact shadows under him on the seat. The remote rests in his hand at normal household size with real weight. No floating props. LIGHTING Window key from frame-left, warm and fading toward evening, dimmer as it goes. Cool flatscreen glow as fill, its hue shifting with the content of @video_1, rimming the back of his hair and shoulders. The screen surface catches a soft reflection of the room and window. White balance fixed at 5200K. Light atmospheric haze softening the depth between camera and TV. SCREEN REALISM The TV picture reads as a real physical panel being filmed, not a clean digital overlay: the picture sits behind the glass panel and carries a faint glass sheen and soft glare, dim reflections of the room and window on the screen surface, a subtle pixel/sub-pixel grid and gentle moiré, slight bloom on the brightest areas, mild panel contrast and color shift, and a small off-axis perspective from the side viewing angle. The underlying picture stays fully recognizable as @video_1. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. The native audio of @video_1 from the TV, plus quiet room tone. No subtitles, no added voiceover. STYLE 8K photorealistic live-action cinematic, real lens capture, naturalistic master cinematography. 24fps, real-time, smooth motion, no jitter, 16:9, total duration 6 seconds. POSITIVE LOCKS The TV picture matches @video_1 1:1 in content, timing and color, fully recognizable throughout — while looking like a real filmed panel with glass sheen, soft glare, faint room reflections, a subtle pixel grid and gentle moiré, mild bloom and a slight off-axis angle. The flatscreen stays fully visible and unobstructed — the man's head stays low and to one side. He leans forward toward the TV and stays off the backrest the entire time, seated on the front edge of the sofa. The remote @REMOTE stays resting in his hand, normal household size, not pressed. The camera stays fully static and locked off, one continuous shot, no cut, no move. The flatscreen keeps the exact same size and position as in the @ADIL'S-ROOM room the entire time — it does not grow, shrink, or move. Screen content stays unbranded — no logos or UI beyond what is already in @video_1. Same @ADIL identity and wardrobe throughout. Contact shadows under him read clearly. No eye glow.

Perfect, now let's stitch it all together and see how scene 5 came out完美,现在把它们拼接在一起,看看场景 5 的效果。


Scene 6 - The close-up on the remote场景 6 —— 遥控器特写

A short one: a close-up of the hand pressing the channel button. Easy, right? I am going to add a red arrow directly onto the prop sheet pointing to the right button, drop it into Claude, and ask:一个短镜头:手按频道按钮的特写。简单吧? 我会在道具表上直接加个红箭头指向正确的按钮,丢进 Claude,然后问:

Describe exactly where the red arrow is pointing. And specify that the left thumb is pressing right there above the 'CH' button, exactly where the red arrow is pointing.准确描述红箭头指向哪里。并指定左手大拇指正按在‘CH’按钮上方,也就是红箭头指向的地方。

SCENE CONTEXT Extreme close-up of @ADIL's left hand resting on his left thigh, holding @REMOTE, his thumb making one single deliberate channel-up press. The lived-in living room falls into soft blur behind. One continuous static close-up, 4 seconds. ACTIVE REFERENCES @REMOTE — matte black household streaming remote. 100% matches the reference. Face layout top to bottom: chartreuse "higgs 4K" logo and a round power button at top; a round navigation pad with up/down/left/right chevrons below; then a lower row of two rockers — a VOL rocker on the LEFT ("+" top, "VOL" mid, "−" bottom) and a CH rocker on the RIGHT (UP chevron ˄ top, "CH" mid, DOWN chevron ˅ bottom). The reference's red arrow points precisely to the CH-UP chevron (˄) directly above the "CH" label. @ADIL'S-ROOM — STYLE REFERENCE ONLY, not a fixed keyframe; the out-of-focus backdrop carries this room's character. Model extends the world, never reproduced 1:1. LOCATION MAP Foreground (sharp, on the strong third): the left hand and @REMOTE on the thigh, the CH rocker and thumb. Midground: the thigh and lap falling off into softness. Background: the room reduced to a soft out-of-focus wash of warm neutral surfaces and furniture silhouettes from @ADIL'S-ROOM. Window light enters from frame-left; a cool spill comes from off-frame front. Camera anchored on the shadow side. FIRST FRAME / BLOCKING Non-empty opening frame: the left hand already resting on the left thigh holding @REMOTE, thumb resting just over the CH-UP chevron, alive and steady — nothing frozen, faint breathing motion. Rule of thirds: the remote runs on a slight diagonal, CH rocker and thumb on the strong third, tight on hand and remote with the soft room wash behind. FORMAT MODE One continuous static close-up; the camera does not cut and does not move on its own. 4-second clock, one single press only. OPTICS Extreme close-up at 12° FOV, tight on the lower-right of @REMOTE and the thumb — hand and remote crisp, background dissolving into soft out-of-focus blur. Shallow depth of field, prime tele-detail character, 180° shutter motion blur. No drift. CAMERA Locked off on the shadow side, eye-level to the remote, held just above and in front of the hand looking down at the remote face on the thigh, CH rocker and thumb centered. Focus held on the CH-UP chevron and the thumb. No pan, no push, no shake. Wide tonal latitude, soft highlight roll-off on the matte plastic and skin. ACTION 0.0s–1.0s — the left hand rests on the left thigh holding @REMOTE, thumb resting on the lower-right face just over the CH-UP chevron (˄) directly above the "CH" label, where the red arrow points; matte black surface catching soft cool ambient light. 1.0s–2.0s — the thumb presses the CH-UP chevron (˄) one single time — a deliberate press, the button giving slightly under the pad, one soft mechanical click as the channel changes. The thumb stays clear of the VOL rocker to the left, the navigation pad above, and the CH-DOWN chevron (˅) below — only CH-UP. 2.0s–4.0s — the thumb lifts off and settles back, the hand resting steady on the thigh, @REMOTE held still, no further presses. PERFORMANCE One deliberate, unhurried thumb press — settles on the button, presses down once with a small give, releases. Pore-level realism on the hand: vellus hair, knuckle creases, nail detail, capillary flush, pore-shadows matching the soft window and ambient light. PHYSICS @REMOTE rests with real weight in the grip on the thigh; the thumb pad compresses the CH-UP chevron once with a tiny travel and release; the hand stays steady through the press. No floating props. LIGHTING Soft window daylight from frame-left as warm key, fading toward evening; a cool off-frame-front spill as fill, rimming the knuckles and the top edge of the remote. Faint atmospheric haze. WB locked 5200K. The remote stays plain matte plastic lit only by ambient light — surface and buttons unlit, no glow anywhere. COLOR GRADE Warm room neutrals on skin and matte black plastic as the dominant material, the cool off-frame spill as the secondary beam catching the knuckles and remote edge, and a chartreuse #d1ef17 accent living only on the "higgs 4K" logo — color tied to surface and light, not a flat list. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. One single soft mechanical button click on the press, quiet room tone. STYLE 8K photorealistic live-action macro, naturalistic real-lens capture, fine grain. 24fps, real-time, smooth motion. OUTPUT SETTINGS 16:9. Real-time, total duration 4 seconds. POSITIVE LOCKS It is the LEFT hand on the LEFT thigh holding @REMOTE, the backdrop a soft out-of-focus wash of @ADIL'S-ROOM. The thumb presses the CH-UP chevron (˄) directly above "CH" on the lower-right CH rocker, where the reference red arrow points — exactly ONE time, a single clear press-and-release, then it lifts off and the hand holds still. @REMOTE stays matte black with the chartreuse "higgs 4K" logo, lit only by ambient light, surface and buttons unlit with no glow. Framing stays tight on the hand and CH rocker. @REMOTE stays normal household size. Static locked-off close-up, 4 seconds, real-time, no cut, no move. Eyes/skin natural — no glow.

Alright, the tip worked, the shot's good, we lock it and move on to scene 7.好了,这招奏效了,镜头很棒,锁定它,进入场景 7。


Scene 7 - Flipping the channels场景 7 —— 换台

For this, I just need to be sitting there flipping through the TV channels. The main trick here is that the TV needs to be playing 4K footage, but Seedance has to generate that on-screen all by itself. Let's get the prompt. So I'm typing to Claude:这个镜头,我只需要坐在那换台。这里的诀窍是电视里需要播放 4K 画面,但 Seedance 必须自己把画面生成出来。 我们来写提示词。我给 Claude 打字:

Write a prompt for the next scene. The main character is sitting on the couch watching the news, seen from behind, the TV fully in frame. The broadcaster on screen says that Seedance 4K is unreal. Then the hero clicks through the channels three times using his remote and lands on different ads each time. Hard cut to his face: surprise, he says "Wow." First shot static. 10 seconds total.为下一个场景写个提示词。主角坐在沙发上看新闻,背影,电视完全在画面内。屏幕上的播音员说 Seedance 4K 不可思议。然后英雄用遥控器连按三次频道,每次都跳到不同的广告。硬切到他的脸:惊讶,他说“哇”。第一个镜头静态。总共 10 秒。

And to make the transition logical, let's have the final channel be the opening shot of our next scene. We won't jump ahead to generating the whole scene - I'll just make the location with the same workflow and drop it into Claude's prompt as an asset, to save time. I wanted the next scene to be a super epic battle with wide shot, so I need a dark-fantasy location.为了让转场逻辑通顺,让最后一个频道正好是我们下一个场景的开场镜头。 我们不急着生成整个场景——为了省时间,我用同样的流程制作场景,并作为素材丢进 Claude 的提示词里。我希望下一个场景是一场史诗般的广角战斗,所以我需要一个黑暗奇幻场景。

Vast open grassy battlefield at the foot of towering rocky mountains, low ground-level wide shot with dewy green grass filling the foreground. The plain is completely empty — no soldiers, no armies, no people, just open windswept grassland. On the jagged mountain ridge in the background stands a colossal white-and-gold elven fortress with soaring slender spires and gothic arched bridges carved into the stone. Soft morning sun sits low and partly hidden behind the peaks on the right, giving gentle warm rim light and subtle haze without strong glare. Deep blue sky with scattered clouds, golden-hour backlight, atmospheric depth, drifting mist over the land. Epic fantasy, photorealistic, cinematic.耸立的岩石山脚下广阔的草地战场,地面广角镜头,前景是沾满露水的绿草。平原完全空旷——没有士兵,没有军队,没有人,只有广阔的被风吹拂的草地。背景的锯齿状山脊上矗立着一座巨大的白金精灵堡垒,高耸的细长尖塔和哥特式拱桥雕刻在岩石中。清晨的太阳低垂,部分被右侧的山峰遮挡,提供柔和的暖色轮廓光和微妙的雾霾,没有强烈的眩光。深蓝色天空,云层稀疏,黄金时刻逆光,大气深度,大地上漂浮着薄雾。史诗奇幻,照片级真实感,电影感。

Let's run with this one. Now I am attaching this location so that it becomes the final shot playing on the TV, and I am updating the prompt from our previous generation:就用这个。 现在我附上这个场景,让它成为电视里播放的最后一个镜头,并更新我们之前生成的提示词:

let's specify what's on the TV: after the news anchor, first channel is a perfume commercial, second is a car commercial - both with no text or logos on screen. Third click switches to @Image 4. Then the cut to his reaction.指定电视里播放的内容:新闻主播之后,第一个频道是香水广告,第二个是汽车广告——屏幕上都没有文字或 Logo。第三次点击切换到 @Image 4。然后切到他的反应。

SCENE CONTEXT Lived-in living room toward evening. @ADIL sits leaning forward on the front edge of the sofa, watching a small flatscreen across the room, @REMOTE in his left hand. He flips channels — news, two ads, then a sweeping fantasy film — and reacts. The shot is always the living room: a small ordinary TV across the room plays everything as a little inset picture. Two shots, one hard cut, 10-second clock. ACTIVE REFERENCES @ADIL — the man. 100% matches the reference. Shot 1: from behind, leaning forward on the front edge, back/shoulders/back of head to camera, face turned away, low and to one side so the small TV stays fully visible, @REMOTE in his LEFT hand. Shot 2: face revealed in a medium reaction, LEFT arm extended with @REMOTE pointed at the TV, then lowering it. Same wardrobe throughout. @REMOTE — normal household TV remote in his LEFT hand, chartreuse standby LED, pointed at the TV through the presses. 100% matches the reference. @ADIL'S-ROOM — STYLE REFERENCE ONLY, not a fixed keyframe; model extends the world freely, never reproduced 1:1. @IMAGE4 — the fantasy film shown inside the little TV: a white, gold-spired castle on a misty grey mountain ridge above a vast green meadow at golden sunrise, drifting mist. 100% matches the reference, displayed only inside the small TV rectangle. LOCATION MAP The frame is always the living room, camera behind the sofa. Foreground (low, frame-left): the back of @ADIL's head and shoulders, full-size, rim-lit by the screen, @REMOTE in his left hand. Midground: the sofa front edge and the floor across the room. Background (upper-middle, ~3–4 m away): a small ordinary flatscreen TV filling roughly a quarter (~25%) of the frame, framed by its black bezel and the wall around it; the side window frame-left spilling fading daylight. Everything that plays stays a small picture inside that little TV rectangle — the room and the back of his head stay full-size around it. Camera on the shadow side. Shot 2 re-sets in front on his face. FIRST FRAME / BLOCKING Non-empty opening frame: the little distant TV already playing the news, @ADIL already seated leaning forward, head low in the foreground, breathing and faintly shifting — nothing frozen. Rule of thirds: his head sits low and to one side so the small flatscreen stays fully visible in the upper-middle background. The whole room is in view. FORMAT MODE Timed multishot, exactly ONE HARD CUT at 0:07; the camera otherwise holds still behind the sofa with the whole room in view. 10-second clock. OPTICS SHOT 1 (0:00–0:07): medium-wide at 47° FOV, @ADIL's head and shoulders full-size low in the foreground, the small flatscreen (~25% of frame) unobstructed in the upper-middle background. SHOT 2 (0:07–0:10): medium at 29° FOV, eye-level on his face, his extended LEFT arm and @REMOTE in frame. Real cine-lens character, 180° shutter motion blur. No drift mid-segment. CAMERA SHOT 1: fully static, locked off, positioned BEHIND the sofa on the shadow side, looking over @ADIL's lowered head across the room toward the small distant flatscreen, the whole room held in view, his face turned away — no move. SHOT 2: static, locked off, re-set in front at eye level on his face, the extended arm and remote in frame — no move. Wide tonal latitude, soft highlight roll-off holding the bright little screen. ACTION 0:00–0:02 — SHOT 1: on the small distant flatscreen the news plays — a single anchor at a clean studio desk speaking straight to camera, saying in English "Seedance 4K is unreal." @ADIL sits leaning forward, watching, cool screen glow rimming his hair; the room fills the frame, the TV stays small in the background. 0:02–0:04 — still SHOT 1: he raises his LEFT hand with @REMOTE and presses — soft click — the little screen changes to Advertisement 1: a clear faceted glass perfume bottle of golden liquid slowly rotating on a mirrored surface, light refracting into soft rainbow glints, a gentle puff of mist; the room glow turns warm gold. 0:04–0:06 — he presses again — soft click — the little screen changes to Advertisement 2: a sleek silver sports car with a smooth seamless polished body gliding along a coastal cliff road at golden hour, sea glittering below, a thin dust trail behind the tyres; the glow shifts cool blue. 0:06–0:07 — he presses a third time — soft click — the little screen changes to @IMAGE4: the castle on the misty ridge above the green meadow at sunrise, staying small inside the distant TV, warm golden glow washing the room. The room and his head stay full-size around the small TV. 0:07 HARD CUT. 0:07–0:10 — SHOT 2: medium on his face, eye-level, his LEFT arm still extended with @REMOTE pointed at the TV. A micro-pause, then moderate surprise — eyebrows lift, eyes widen a touch, a small open-mouth beat — and he says "WOW," lowering his LEFT arm on the same beat as the word. Restrained and human. PERFORMANCE Shot 1 (from behind): deliberate, unhurried raise of the arm and presses, body settled on the front edge. Shot 2: a held micro-pause before the reaction, eyebrows lifting, eyes widening a touch, lips parting for a single "Wow" — moderate, genuine surprise, never cartoonish. Pore-level realism: vellus hair, asymmetric moles, capillary flush, wet living eyes with screen catch-lights, visible breath and chest rise. PHYSICS Real body weight on the front edge of the cushions, the cushions compressing where he sits, clear contact shadows under him on the seat. @REMOTE carries real weight in his left hand; each press gives a tiny travel and release. No floating props. LIGHTING Window key from frame-left, warm and fading toward evening, dimmer as it goes; camera on the shadow side; faint atmospheric haze. Cool flatscreen glow as fill, its hue shifting with each channel — warm gold on the perfume, cool blue on the car, warm gold on @IMAGE4 — rimming the back of his hair and shoulders. In Shot 2 the screen glow lights his face from frame-front. WB locked 5200K. COLOR GRADE Warm fading-daylight room neutrals as the dominant material, the cool TV-screen glow as the secondary beam (shifting hue per channel), and a chartreuse #d1ef17 accent living only on the @REMOTE standby LED and a flicker of the screen glow — color tied to source and surface, not a flat list. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. News anchor (from the TV, English): "Seedance 4K is unreal." Then a few soft channel-change clicks with abrupt TV-audio shifts; soft ambient wind atmosphere as @IMAGE4 comes up. @ADIL (Shot 2, English): "Wow." Room tone. No subtitles. STYLE 8K photorealistic live-action — no 3D render, no game engine, no game-cutscene aesthetic. Naturalistic master cinematography, fine film grain. OUTPUT SETTINGS 16:9. Real-time throughout. Total duration 10 seconds. POSITIVE LOCKS The frame is always the living room, camera behind the sofa with the whole room in view. The TV is a small ordinary television filling about a quarter of the frame in the upper-middle background; every channel — news, both ads, and @IMAGE4 — appears as a small picture inside the distant TV's bezel, the room and the back of @ADIL's head full-size around it. Exactly one HARD CUT, at 0:07; otherwise the camera holds still. @ADIL leans forward on the front edge and uses his LEFT hand with @REMOTE pointed at the TV through three unhurried switches: news → perfume ad → sports-car ad → @IMAGE4. His head stays low and to one side so the small TV stays fully visible. Shot 2 reveals his face for the "WOW" with his LEFT arm extended, then he lowers it on the same beat as the word — moderate, natural, restrained surprise. Same @ADIL identity and wardrobe across the cut. TV and @REMOTE stay normal household size. Contact shadows under him on the sofa read clearly. Eyes stay natural — no eye glow.

Now everything plays perfectly inside the television, the image is deep and realistic, and the last frame lands exactly where the next scene begins.现在一切都在电视里完美播放,画面深邃真实,最后一帧正好落在下一个场景开始的地方。


Scene 8 - The epic battle场景 8 —— 史诗战斗

The location was ready from Scene 7, so we built the characters: an elf, an orc, and two tiny fighters for the ground level - a flower-headed plant warrior and a troll. 场景 7 的场景已经准备好了,所以我们做了角色:一个精灵、一个兽人,还有两个地面上的小战士——一个花头植物战士和一个巨魔。

Cinematic close-up portrait of a massive brutal orc warrior leaning forward toward the camera on a plain solid neutral grey background. Huge hulking build, thick grey-blue weathered scarred skin, heavy brow, small glowing amber eyes, broad flat nose, large protruding lower tusks and jagged teeth, pointed ears, dark crude tribal markings carved into the forehead. He wears heavy battered dark iron and blackened steel plate armor with rough overlapping scales, chainmail and worn leather straps. Menacing, intimidating, hunched aggressive posture. Ultra-realistic cinematic photograph, soft dramatic lighting, shallow depth of field, plain solid neutral grey background, photorealistic, highly detailed skin and metal textures, no text, no logo.巨大的残暴兽人战士的电影感特写肖像,在纯色中性灰背景上向前倾斜。体型巨大,厚实的蓝灰色风化疤痕皮肤,沉重的眉骨,发光的琥珀色小眼睛,宽扁的鼻子,巨大的下突长牙和锯齿状牙齿,尖耳朵,额头上刻着粗犷的部落标记。穿着厚重的破损铁质和黑钢板甲,带有粗糙的重叠鳞片、锁子甲和磨损的皮革带。威胁感,恐吓感,前倾的攻击姿态。超写实电影感摄影,柔和的戏剧性光线,浅景深,纯色中性灰背景,照片级真实感,高度细致的皮肤和金属纹理,无文字,无 Logo。
Cinematic close-up portrait of a massive brutal orc warrior leaning forward toward the camera on a plain solid neutral grey background. Huge hulking build, thick grey-blue weathered scarred skin, heavy brow, small glowing amber eyes, broad flat nose, large protruding lower tusks and jagged teeth, pointed ears, dark crude tribal markings carved into the forehead. He wears heavy battered dark iron and blackened steel plate armor with rough overlapping scales, chainmail and worn leather straps. Menacing, intimidating, hunched aggressive posture. Ultra-realistic cinematic photograph, soft dramatic lighting, shallow depth of field, plain solid neutral grey background, photorealistic, highly detailed skin and metal textures, no text, no logo.
Covered muscular body, glowing bioluminescent veins beneath the bark, smooth wooden mask with glowing amber eyes, massive orange protea flower petals forming a majestic mane around the head, layered leaf armor made from real rainforest foliage, clawed hands, digitigrade predator feet, wielding a massive living nature-forged sword, blade carved from ancient blackwood and hardened amber resin, glowing sap flowing through engraved organic channels in the blade, root-like guard and vine-wrapped handle, highly detailed organic weapon design, realistic plant anatomy, photorealistic creature design, cinematic realism, neutral gray studio background, soft studio lighting, symmetrical presentation, creature concept reference sheet, full body character turnaround on the left, detailed facial portrait on the right, ultra detailed bark, leaves and flower textures, movie-quality practical effects creature, Weta Workshop realism, sharp focus, HDR, 8k覆盖肌肉的身体,树皮下发光的生物发光静脉,带有发光琥珀色眼睛的光滑木制面具,巨大的橙色帝王花花瓣在头部周围形成威严的鬃毛,由真正的雨林叶片制成的层叠叶甲,爪状手,趾行式捕食者脚,挥舞着一把巨大的活体自然锻造剑,刀刃由古老的黑木和硬化琥珀树脂雕刻而成,发光的树液流过刀刃上雕刻的有机通道,根状护手和藤蔓包裹的手柄,高度细致的有机武器设计,真实的植物解剖结构,照片级真实感生物设计,电影级真实感,中性灰摄影棚背景,柔和的摄影棚光,对称展示,生物概念参考表,左侧全身角色转面,右侧详细面部肖像,超细节的树皮、叶子和花朵纹理,电影级实用特效生物,维塔工作室真实感,锐利对焦,HDR,8K
Character design sheet, single subject shown twice on one image, split into two panels on a seamless neutral grey studio background. LEFT PANEL — full body, head-to-toe, heavy hunched aggressive stance: A massive brutish humanoid swamp warrior with thick, leathery dark-green and olive hide, slick and wet, streaked with mud, silt and patches of slimy bog moss and algae. Heavy hulking anatomy — broad shoulders, thick arms, a wide muscular torso — the skin mottled with murky brown and swamp-green tones, clinging duckweed and damp sheen. Curved yellowed tusks jut from the lower jaw, bony ridges and warty knobs run along the shoulders and forearms. Crude armor of rusted iron plates, bone fragments and weathered leather straps lashed to the body with frayed rope, caked in dried mud. Webbed clawed hands and heavy splayed feet. In one hand he grips a heavy crude greatsword of pitted rusted iron, the blade nicked and stained, the grip bound in damp leather and swamp vine. He stands alone on a clean empty grey floor with nothing around his feet. RIGHT PANEL — close-up portrait of the same face, three-quarter angle: A brutal amphibian-like face with thick mottled dark-green and brown hide, a wide flat heavy brow, broad nostrils and a massive jaw with protruding yellowed tusks. Wet slimy skin streaked with mud and clinging algae, small leeches and bog moss in the deep skin folds. Small sunken eyes glowing a dull murky amber-green. Instead of hair, tangled wet reeds, swamp weeds and bony spines drape back over the skull. Drips of water and slime on the surface. STYLE: hyper-detailed 3D character render, soft moody diffused lighting, gentle shadows, rich swamp-green and muddy-brown palette with rusted iron accents, wet slimy skin with subsurface sheen, photorealistic textures, neutral dark-grey gradient background, clean empty floor, no props on the ground, no text, no logo.角色设计表,单一主体在一张图上出现两次,分为两个面板 在无缝中性灰摄影棚背景上。 左侧面板 — 全身,从头到脚,沉重的弓背攻击姿态: 一个巨大的粗野人形沼泽战士,皮肤厚实,呈深绿色和橄榄色, 湿滑,布满泥浆、淤泥和成片的粘稠沼泽苔藓和藻类。沉重 魁梧的解剖结构 — 宽肩膀,粗手臂,宽阔的肌肉躯干 — 皮肤斑驳 带有深棕色和沼泽绿色调,粘着浮萍和潮湿的光泽。弯曲的黄色 长牙从下颚突出,骨质脊和疣状结节沿着肩膀和 前臂分布。粗糙的锈铁板、骨碎片和风化的皮革带 用磨损的绳子绑在身上,结满干泥。有蹼的爪子和沉重的 外翻脚。一只手里抓着一把沉重的粗制大剑,由坑洼的锈铁制成, 刀刃有缺口和污渍,握把绑着潮湿的皮革和沼泽藤蔓。他独自站立 在干净的空灰色地板上,脚边什么都没有。 右侧面板 — 同一张脸的特写肖像,3/4 角度: 一张残酷的类似两栖动物的脸,皮肤厚实斑驳,呈深绿色和棕色,宽平的 沉重眉骨,宽大的鼻孔和巨大的下颚,突出的黄色长牙。湿润粘稠的 皮肤上布满泥浆和粘着的藻类,深层皮肤褶皱里有小水蛭和沼泽苔藓。 深陷的小眼睛发出暗淡的浑浊琥珀绿色光。头上没有头发,而是纠缠着湿润的 芦苇、沼泽杂草和骨刺。表面有水滴和粘液。 风格:超细节 3D 角色渲染,柔和的情绪化漫射光,柔和的阴影, 丰富的沼泽绿和泥棕色调,带有锈铁点缀,带有次表面光泽的湿润粘稠皮肤,照片级真实感纹理,中性深灰渐变背景,干净 空地板,地上没有道具,没有文字,没有 Logo。

Now, let’s load all the assets into Claude - the location, the elf, the monsters—and ask for the prompt.现在,把所有素材——场景、精灵、怪物——加载到 Claude,然后要求生成提示词。

"Write me a prompt for a fantasy video. There should be a massive battle between elf troops and an orc army, and then the camera slowly dives downward, revealing a flower-warrior and a troll dueling with weapons."“给我写一个奇幻视频的提示词。画面应该是一场精灵军队和兽人大军之间的宏大战斗,然后镜头缓慢向下俯冲,露出一个花战士和一个巨魔正在用武器决斗。”

SCENE CONTEXT One continuous, fast and dynamic FPV-drone battle shot with a dramatic scale shift. An elven army charges in from the LEFT with swords; an orc horde charges to meet it from the RIGHT, every clash pairing an elf against an orc. The drone flies fast through the middle of the clash, weaving between the fighting bodies, then dives down between TALL blades of grass to ground level — where two tiny, palm-sized fighters duel one-on-one while the army battle rages above. ACTIVE REFERENCES @castle — green grass field rising to the white-and-gold spired elven citadel on a misty grey mountain, low dawn sun flaring screen-right, soft mist drifting down the slopes. 100% matches the reference (STYLE REFERENCE ONLY, model extends the world). @elf — STYLE reference for the elven army: silver-and-gold leaf-filigree armor, pale-blue capes, elven swords. The army is a varied host — men and women, each a distinct face, hair color, build, helmet and cape, every elf unique while keeping this armor-and-weapon style, each fighting with a sword and striking only at orcs. 100% matches the reference style. @orc — the army orcs: dry, upright SOLDIER orcs in battered black-iron plate, green-grey skin, tusks, amber eyes, heavy iron weapons — ordinary orc-sized, lean and brutish, about a head taller than the elves, each unique, striking only at elves, staying in the large army up top. 100% matches the reference. @flower-warrior — forest knight: carved wooden face with glowing amber eyes, orange flower-petal mane, green leaf mantle, bark skin, vine-wrapped spear and a glowing amber-cored shield. 100% matches the reference. @troll — BOG TROLL: a hulking, hunched swamp troll, thick green hide caked in wet moss, slick algae and dripping mud, long trailing swamp-weed hair, big jutting tusks, small amber eyes, ragged hide-and-scrap armor and a rope harness, gripping a heavy pitted rusty sword. A wet, mossy, primitive swamp creature — a troll, not an orc, clearly different from the dry black-iron soldiers above. 100% matches the reference. LOCATION MAP The world of @castle. Background: the white-gold spired citadel on the misty grey mountain, dawn sun flaring screen-right, mist down the slopes, bright battle bokeh up high. Midground: the two charging lines colliding across the open grass field — elves from screen-LEFT, orcs from screen-RIGHT. Foreground (ground beat): TALL wet grass blades, soil, dew, with the normal-scale soldiers' legs, feet and boots towering and crashing down like giants around the two palm-sized fighters. The drone path runs side-on along the lines → through the clash → down between the grass to the soil. SCALE LOCK Orcs and elves are ordinary human-scale warriors. @flower-warrior and @troll are TINY — each about the size of a human palm, standing as tall as a single blade of grass on the soil. Because they are so small, the normal-sized elves and orcs rise above them like giants, so at ground level the soldiers' legs, feet and boots fill the frame and crash down beside the two tiny fighters. FIRST FRAME / BLOCKING Non-empty opening frame: a wide SIDE view, both armies already charging and closing — elven swordsmen rushing in from screen-LEFT, orcs from screen-RIGHT, the two lines about to collide across @castle, citadel and screen-right sun behind. Everything already in motion, nothing frozen. FORMAT MODE One unbroken continuous take; a fast, free-flying FPV-drone camera holds the whole shot, the camera does not cut on its own. OPTICS Wide ~84° up high for scale and speed, tightening to ~47° as it weaves through the fighters, then macro ~18° at ground level among the grass. Deep focus up high, shallow macro focus down low. Strong motion blur on the fast moves. No drift mid-beat. CAMERA Aggressive, weightless, dynamic FPV-drone movement, always moving. Beat 1 — open on the wide side view: the two charging lines rushing together across @castle. Beat 2 — the drone flies fast and low THROUGH the middle of the clash, weaving and banking between the colliding elves and orcs, slipping past swinging swords. Beat 3 — it whips downward and punches between the TALL grass blades to the soil, threading dynamically around the two palm-sized fighters while overhead the legs and boots of the normal-sized soldiers stomp and crash around the grass. Continuous, fluid, energetic. ACTION Up top the armies collide — elf against orc all across the line: each elven swordsman duels an orc, each orc swings at an elf, same-side allies fighting shoulder to shoulder against the enemy. As the drone reaches the grass, the two tiny fighters duel ONE-ON-ONE, fast and dynamic: @flower-warrior lunging and thrusting the vine spear, @troll parrying and counter-swinging the heavy rusty sword — both dodging, spinning, blades clashing, leaves and embers scattering, mud and water flicking off the troll, while orc and elf feet pound the earth nearby, just missing them, dust and dew bursting with each footfall. PERFORMANCE Up top: ferocious committed combat, varied unique faces straining, war-cry mouths, weighted swings. At ground level: @flower-warrior moves fast and agile, light on its feet; @troll moves with heavy lumbering mass, shoulders rolling, a slow brute power against the knight's speed — a clear contrast of styles, both reading as alive and reactive. PHYSICS Realistic momentum and weight — soldier footfalls send tremors and burst dust and dew at macro scale; the tall grass whips and bends in the drone's wake; @flower-warrior is fast and agile, @troll heavy and lumbering with water and mud dripping and flicking from its hide, both with weighty creak and squelch. Every pose flows continuously into the next, motion fully connected, no floaty movement. LIGHTING Cinematic dawn — warm low backlight and sun flare screen-right from @castle, soft mountain mist, volumetric light and atmospheric haze up high with bright battle bokeh. At ground level sun glints through the tall wet grass; the soldiers' legs cast moving shadows over the tiny fighters; @flower-warrior's amber core-glow and the wet sheen on @troll's mossy hide read against the shadowed soil. COLOR GRADE Warm dawn gold from the low backlit sun as the dominant beam on the grass and armor, cool misty grey-blue mountain atmosphere as the secondary tone, with accents of pale-blue elven capes, silver-gold filigree, the orange petal mane, amber shield-glow and the wet green moss of the troll — color tied to material and the dawn light, glowing and filmic, not a flat list. STYLE 8K photoreal cinematic, ultra-detailed, high-budget epic fantasy film look — natural skin, bark, hide, moss, metal and leaf textures with fine micro-detail, macro-photography texture on grass and soil, real organic grain, filmic grade. OUTPUT SETTINGS 16:9. Real-time, one continuous take. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. Roaring charge, war cries, clashing steel and sword rings, thudding footsteps overhead; up close the fast sharp clashes of the two tiny weapons, wet squelching, dripping water, creaking wood, scattering leaves, bursting dew, wind through tall grass. POSITIVE LOCKS One continuous fast FPV-drone take — opens with elves charging from screen-LEFT and orcs from screen-RIGHT, flies dynamically THROUGH the battle, then dives down between TALL grass blades to ground level. Orcs and elves stay ordinary human-scale warriors. Every clash is elf versus orc — elves strike at orcs, orcs strike at elves, same-side allies fight side by side against the enemy line. The elven army follows the @elf style as a varied host — men and women, each a distinct face, hair and armor — every elf fighting with a sword in hand. At ground level there are exactly TWO tiny fighters: @flower-warrior versus @troll. The big army above is dry black-iron orc soldiers (@orc); the tiny ground fighter @troll is the moss-and-mud bog troll with a rusty sword — a separate creature, a troll and not an orc. Each orc and elf unique. Dynamic motion throughout, crisp stable readable picture, temporally consistent frames.

All the elves are identical. Let's make a character sheet with a few different elves and load that image instead of the single one.所有的精灵都长得一样。让我们做一个包含几个不同精灵的角色表,加载那张图而不是单人的。

Full-body character lineup of four different elf warriors standing side by side in a row on a seamless neutral grey studio background, soft diffused studio lighting with gentle shadows, ultra-realistic, photographic, cinematic quality, in the visual style of high fantasy film production stills. All four have authentic, refined elven faces — ethereal and noble, with high sharp cheekbones, fine straight noses, smooth flawless pale skin, almond-shaped eyes, delicately pointed ears and serene, serious, dignified expressions. Faces look like real beautiful people photographed on camera, not stylized or cartoonish — subtle, regal, otherworldly beauty, never exaggerated or comical. Long, sleek hair, finely braided in elegant elven style. Each elf is shown head-to-toe standing naturally and relaxed, arms resting at their sides, facing forward with a calm neutral expression — no dramatic posing. Tall, slender, graceful proportions. Every elf is fully clad in elven MILITARY combat armor — full layered plate covering the chest, shoulders, pauldrons, arms, gauntlets, thighs and greaves, sculpted with flowing organic engravings, leaf and feather motifs, fitted and battle-practical, worn over a fitted gambeson rather than loose dresses or robes; a modest cloak may hang from the shoulders. No bows, no arrows, no quivers — each is armed only with a slender elven sword, simply sheathed at the hip or held loosely at the side. The armor is refined and restrained, in light, soft, muted tones, not bright or garish. Each character has a slightly different armor design and a distinct subtle light palette. ELF 1 — female: full polished silver-and-pearl elven plate armor with feather-engraved pauldrons and armored greaves, a short pale ice-blue cloak, platinum-blonde braided hair, a slim longsword sheathed at her hip. ELF 2 — male: full champagne-gold and ivory elven plate armor with delicate leaf etching, armored arms and legs, a muted sage-green cloak, long ash-brown hair, an ornate curved sword at his side. ELF 3 — female: full rose-gold and cream elven plate armor with fine filigree, fitted breastplate and leg plates, a light dove-grey half-cloak, soft auburn hair in braids, a slender straight sword sheathed at her hip. ELF 4 — male: full pale silver and light steel-blue elven plate armor with subtle engraving, heavy pauldrons and greaves, a soft pale-grey cape, long silver-white hair, a broad elegant sword at his side. STYLE: ultra-realistic cinematic photograph, true-to-life elven faces, natural skin texture, real polished metal plate armor and fine engraved detail, full elven military combat armor, light restrained palette with subtle variation, natural relaxed standing poses, soft even lighting, clean neutral grey gradient background, full body visible head to toe, shot on a professional camera, photorealistic, highly detailed, no bows, swords only, no exaggerated cartoonish features, no text, no logo.四个不同精灵战士的全身角色阵容,并排站成一排 在无缝中性灰摄影棚背景上,柔和的漫射摄影棚光,带有柔和的 阴影,超写实,摄影级,电影质量,高奇幻电影制作剧照的视觉风格。 四个人都有地道、精致的精灵面孔——空灵而高贵,高而尖的 颧骨,挺直的鼻子,光滑无瑕的苍白皮肤,杏仁眼,精致的 尖耳朵,表情宁静、严肃、庄重。面孔看起来像是在摄像机前拍摄的真实美丽 的人,而不是风格化或卡通化——微妙、高贵、超凡脱俗的 美,绝不夸张或滑稽。长而顺滑的头发,编成优雅的精灵 风格。 每个精灵都从头到脚自然放松地站立,手臂垂在身体两侧, 面向前方,表情平静中性——没有戏剧性的姿势。高挑、纤细、优雅的 比例。每个精灵都穿着精灵军事战斗盔甲——全套层叠板甲 覆盖胸部、肩膀、护肩、手臂、护手、大腿和护胫,雕刻着 流动的有机雕刻、叶子和羽毛图案,合身且适合战斗,穿在 合身的衬垫衣外面,而不是宽松的裙子或长袍;肩膀上可能挂着一件朴素的斗篷。 没有弓,没有箭,没有箭袋——每个人只装备一把细长的精灵剑, 简单地收在腰间或松松地握在身侧。盔甲精致而克制, 色调柔和、低调,不鲜艳刺眼。每个角色都有略微不同的 盔甲设计和独特的微妙浅色调。 精灵 1 — 女性:全套抛光银色和珍珠色精灵板甲,带有羽毛雕刻的 护肩和护胫,一件浅冰蓝色短斗篷,白金色的编织长发, 腰间收着一把细长剑。 精灵 2 — 男性:全套香槟金和象牙色精灵板甲,带有精致的叶片蚀刻, 护臂和护腿,一件柔和的鼠尾草绿斗篷,长灰棕色头发,腰间一把华丽的弯剑。 精灵 3 — 女性:全套玫瑰金和奶油色精灵板甲,带有精细的金属丝工艺,合身的 胸甲和腿甲,一件浅鸽灰色半斗篷,编织的柔和赤褐色头发,腰间收着一把细直剑。 精灵 4 — 男性:全套浅银色和浅钢蓝色精灵板甲,带有微妙的 雕刻,沉重的护肩和护胫,一件柔和的浅灰色斗篷,长银白色头发,腰间一把优雅的宽剑。 风格:超写实电影感摄影,真实的精灵面孔,自然的皮肤纹理, 真实的抛光金属板甲和精细的雕刻细节,全套精灵军事战斗盔甲, 明亮克制的色调,微妙的变化,自然的放松站姿,柔和均匀的 光线,干净的中性灰渐变背景,全身从头到脚可见,专业摄像机拍摄,照片级真实感,高度细致,没有弓,只有剑,没有夸张的 卡通特征,没有文字,没有 Logo。
SCENE CONTEXT One continuous, fast and dynamic FPV-drone battle shot with a dramatic scale shift. An elven army charges in from the LEFT with swords; an orc horde charges to meet it from the RIGHT, every clash pairing an elf against an orc. The drone flies fast through the middle of the clash, weaving between the fighting bodies, then dives down between TALL blades of grass to ground level — where two tiny, palm-sized fighters duel one-on-one while the army battle rages above. ACTIVE REFERENCES @castle — green grass field rising to the white-and-gold spired elven citadel on a misty grey mountain, low dawn sun flaring screen-right, soft mist drifting down the slopes. 100% matches the reference (STYLE REFERENCE ONLY, model extends the world). @elf — VARIETY reference for the elven army: a sheet of FOUR different elves — men and women, blonde / dark / auburn / silver hair, silver-, gold- and rose-toned leaf-filigree armor with capes, elven swords. The army is a varied host drawn from these four types — a mix of men and women with different faces, hair colors, builds and armor tones, every elf unique, no two alike, each fighting with a sword and striking only at orcs. @orc — the army orcs: dry, upright SOLDIER orcs in battered black-iron plate, green-grey skin, tusks, amber eyes, heavy iron weapons — ordinary orc-sized, lean and brutish, about a head taller than the elves, each unique, striking only at elves, staying in the large army up top. 100% matches the reference. @flower-warrior — forest knight: carved wooden face with glowing amber eyes, orange flower-petal mane, green leaf mantle, bark skin, vine-wrapped spear and a glowing amber-cored shield. 100% matches the reference. @troll — BOG TROLL: a hulking, hunched swamp troll, thick green hide caked in wet moss, slick algae and dripping mud, long trailing swamp-weed hair, big jutting tusks, small amber eyes, ragged hide-and-scrap armor and a rope harness, gripping a heavy pitted rusty sword. A wet, mossy, primitive swamp creature — a troll, not an orc, clearly different from the dry black-iron soldiers above. 100% matches the reference. LOCATION MAP The world of @castle. Background: the white-gold spired citadel on the misty grey mountain, dawn sun flaring screen-right, mist down the slopes, bright battle bokeh up high. Midground: the two charging lines colliding across the open grass field — elves from screen-LEFT, orcs from screen-RIGHT. Foreground (ground beat): TALL wet grass blades, soil, dew, with the normal-scale soldiers' legs, feet and boots towering and crashing down like giants around the two palm-sized fighters. The drone path runs side-on along the lines → through the clash → down between the grass to the soil. SCALE LOCK Orcs and elves are ordinary human-scale warriors. @flower-warrior and @troll are TINY — each about the size of a human palm, standing as tall as a single blade of grass on the soil. Because they are so small, the normal-sized elves and orcs rise above them like giants, so at ground level the soldiers' legs, feet and boots fill the frame and crash down beside the two tiny fighters. FIRST FRAME / BLOCKING Non-empty opening frame: a wide SIDE view, both armies already charging and closing — elven swordsmen rushing in from screen-LEFT, orcs from screen-RIGHT, the two lines about to collide across @castle, citadel and screen-right sun behind. Everything already in motion, nothing frozen. FORMAT MODE One unbroken continuous take; a fast, free-flying FPV-drone camera holds the whole shot, the camera does not cut on its own. OPTICS Wide ~84° up high for scale and speed, tightening to ~47° as it weaves through the fighters, then macro ~18° at ground level among the grass. Deep focus up high, shallow macro focus down low. Strong motion blur on the fast moves. No drift mid-beat. CAMERA Aggressive, weightless, dynamic FPV-drone movement, always moving. Beat 1 — open on the wide side view: the two charging lines rushing together across @castle. Beat 2 — the drone flies fast and low THROUGH the middle of the clash, weaving and banking between the colliding elves and orcs, slipping past swinging swords. Beat 3 — it whips downward and punches between the TALL grass blades to the soil, threading dynamically around the two palm-sized fighters while overhead the legs and boots of the normal-sized soldiers stomp and crash around the grass. Continuous, fluid, energetic. ACTION Up top the armies collide — elf against orc all across the line: each elven swordsman duels an orc, each orc swings at an elf, same-side allies fighting shoulder to shoulder against the enemy. As the drone reaches the grass, the two tiny fighters duel ONE-ON-ONE, fast and dynamic: @flower-warrior lunging and thrusting the vine spear, @troll parrying and counter-swinging the heavy rusty sword — both dodging, spinning, blades clashing, leaves and embers scattering, mud and water flicking off the troll, while orc and elf feet pound the earth nearby, just missing them, dust and dew bursting with each footfall. PERFORMANCE Up top: ferocious committed combat, varied unique faces straining, war-cry mouths, weighted swings. At ground level: @flower-warrior moves fast and agile, light on its feet; @troll moves with heavy lumbering mass, shoulders rolling, a slow brute power against the knight's speed — a clear contrast of styles, both reading as alive and reactive. PHYSICS Realistic momentum and weight — soldier footfalls send tremors and burst dust and dew at macro scale; the tall grass whips and bends in the drone's wake; @flower-warrior is fast and agile, @troll heavy and lumbering with water and mud dripping and flicking from its hide, both with weighty creak and squelch. Every pose flows continuously into the next, motion fully connected, no floaty movement. LIGHTING Cinematic dawn — warm low backlight and sun flare screen-right from @castle, soft mountain mist, volumetric light and atmospheric haze up high with bright battle bokeh. At ground level sun glints through the tall wet grass; the soldiers' legs cast moving shadows over the tiny fighters; @flower-warrior's amber core-glow and the wet sheen on @troll's mossy hide read against the shadowed soil. COLOR GRADE Warm dawn gold from the low backlit sun as the dominant beam on the grass and armor, cool misty grey-blue mountain atmosphere as the secondary tone, with accents of pale-blue/rose elven capes, silver-gold-rose filigree, the orange petal mane, amber shield-glow and the wet green moss of the troll — color tied to material and the dawn light, glowing and filmic, not a flat list. STYLE 8K photoreal cinematic, ultra-detailed, high-budget epic fantasy film look — natural skin, bark, hide, moss, metal and leaf textures with fine micro-detail, macro-photography texture on grass and soil, real organic grain, filmic grade. OUTPUT SETTINGS 16:9. Real-time, one continuous take. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. Roaring charge, war cries, clashing steel and sword rings, thudding footsteps overhead; up close the fast sharp clashes of the two tiny weapons, wet squelching, dripping water, creaking wood, scattering leaves, bursting dew, wind through tall grass. POSITIVE LOCKS One continuous fast FPV-drone take — opens with elves charging from screen-LEFT and orcs from screen-RIGHT, flies dynamically THROUGH the battle, then dives down between TALL grass blades to ground level. Orcs and elves stay ordinary human-scale warriors. Every clash is elf versus orc — elves strike at orcs, orcs strike at elves, same-side allies fight side by side against the enemy line. The elven army is a VARIED host built from the four elf types in @elf — men and women, mixed hair colors and silver/gold/rose armor tones, every elf unique and no two alike — each fighting with a sword in hand. At ground level there are exactly TWO tiny fighters: @flower-warrior versus @troll. The big army above is dry black-iron orc soldiers (@orc); the tiny ground fighter @troll is the moss-and-mud bog troll with a rusty sword — a separate creature, a troll and not an orc. Each orc and elf unique. Dynamic motion throughout, crisp stable readable picture, temporally consistent frames.

You can see how much smoother the camera got. The movement is super controlled now, it looks like real, professional-level camera work. But the real mind-blower starts when the camera drops down to the ground. Look at that realistic grass texture. In the background the massive battle keeps raging, but the elves and orcs don't turn into a blurry mess, and the image doesn't dissolve into slop. The model completely locks in the geometry of the massive crowd, even with the shifting perspective. Standard 1080p could never hold onto details like this.你可以看到镜头平滑了多少。现在的运动控制得非常好,看起来就像真实、专业的摄像工作。 但真正让人大开眼界的是镜头降到地面的时候。看看那真实的草地纹理。背景中宏大的战斗仍在继续,但精灵和兽人没有变成模糊的一团,画面也没有溶解成烂泥。 模型完全锁定了庞大人群的几何结构,即使视角在变换。 标准 1080p 永远无法保持这样的细节。


Scene 9 - Mom's verdict场景 9 —— 妈妈的评价

The finale: mom walks in, nags him to come eat, glances at the screen - and freezes. "Wait… is that 4K?"结局:妈妈走进来,唠叨他去吃饭,瞥了一眼屏幕——然后愣住了。“等等……那是 4K 吗?”

SCENE CONTEXT Lived-in living room, daytime. @ADIL sits sunk low into the sofa, absorbed in a TV on the wall behind the camera. @MOM walks in from the kitchen, plants herself behind the sofa and nags him to come eat — then her gaze lifts to the screen and irritation turns to surprise. Two shots, one hard cut, 11-second clock. ACTIVE REFERENCES @ADIL — the man. 100% matches the reference. Sunk low into the sofa, centered, body squared to the lens, gaze on the screen just behind the lens, absorbed with a calm natural expression and a faint content half-smile — blinking, breathing, small idle head/hand shifts, fully alive, never frozen, never looking into the lens. @MOM — middle-aged woman, fresh from cooking, drying her hands on a small kitchen towel. 100% matches the reference. Enters through the doorway, takes a few steps, stops behind the sofa, working the towel as she nags. @ADIL'S-ROOM — STYLE REFERENCE ONLY for the room's look and mood, not a fixed keyframe and not the camera angle. Keep the room's textures, palette and clutter; build the geometry from this description, not the input frame. LOCATION MAP The room seen from the TV-wall side, turned around to face the rest of the room (the TV and speakers live on the wall at the camera's back, out of frame). Foreground (lower third, full width): a wide beige sofa with throws, @ADIL sunk low-center on it. Midground: the warm wood floor with a bean bag and clutter. Background: a plain painted back wall carrying a single poster and an open doorway just left of center leading through to an unseen kitchen, where @MOM enters and settles a step behind the sofa back; the side window sits at the RIGHT edge of frame, curtains and string lights, daylight raking in from the right. Generous room depth and headroom behind the sofa. FIRST FRAME / BLOCKING Non-empty opening frame: @ADIL already sunk low-center on the sofa watching, blinking and breathing; @MOM already stepping in through the doorway in the back wall, drying her hands — nothing frozen. Rule of thirds: @ADIL's head on the lower-center third filling the foreground, @MOM small in the upper background behind the sofa, roughly above her son. Low camera at chest height, centered, squared to the room. FORMAT MODE Timed multishot, exactly ONE HARD CUT at 0:07 (right after "now"); the camera otherwise holds still in each shot. 11-second clock. OPTICS SHOT 1 (0:00–0:07): wide two-shot at 63° FOV — @ADIL low-center foreground on the wide sofa, the daylit window at the right edge, @MOM entering through the doorway and settling deep in the background. SHOT 2 (0:07–0:11): medium at 29° FOV, eye-level on @MOM's face and shoulders at the SAME spot behind the sofa. Real cine-lens character, 180° shutter motion blur. No drift mid-segment. CAMERA SHOT 1: static, locked off, low chest-height camera looking across the room toward the sofa and the plain back wall — no move. SHOT 2: static, locked off, re-set tighter on @MOM at the same spot for the reaction as her gaze lifts and drifts forward toward the screen (toward the camera) — no move. Wide tonal latitude, soft highlight roll-off. ACTION 0:00–0:02 — SHOT 1: @ADIL sits sunk into the sofa, absorbed in the screen just behind the lens, calm and natural with a faint content look, blinking and breathing, a small idle shift of head and hand. In the background @MOM steps in through the doorway, drying her hands on the towel, takes a couple of steps and stops behind the sofa, settling her weight. 0:02–0:04 — still SHOT 1: planted behind the sofa, she looks straight down at her son and asks: "What are you watching over there?" @ADIL blinks, only half-listening, eyes still on the screen. 0:04–0:05 — a small beat: she keeps working the towel, eyes on her son, brow tightening; @ADIL shifts a little, blinks, still watching. 0:05–0:07 — still SHOT 1, exasperated, eyes still on him, a little wave of the hand, she finishes scolding: "Come eat — now!" 0:07 — HARD CUT (right after "now"). 0:07–0:08 — SHOT 2: medium on @MOM at the same spot. Her gaze lifts off her son and drifts forward toward the screen behind the camera; the irritation starts to melt, her hands slow on the towel, eyebrows lifting. 0:08–0:10 — eyes widening on the screen, with genuine surprise she asks: "Wait… is that 4K?" 0:10–0:11 — a short half-second beat: she holds the surprised look on the screen, a slow blink and a soft breath, then end. PERFORMANCE @ADIL relaxed and natural, absorbed in the screen — blinking normally, breathing, eyes flicking with what he watches, small idle shifts, a faint pleased look, calm and human rather than frozen, never into the lens. @MOM enters with a real few-step walk and the settle of her weight, exasperated and nagging like a real mother, looking straight down at her son as she scolds; then after the cut a clear shift as her gaze lifts and drifts forward — irritation melts, eyebrows lift, eyes widen, lips part on a surprised question, held briefly. Pore-level realism on both faces: vellus hair, capillary flush, fine lines on @MOM, wet living eyes with catch-lights, visible breath, pore-shadows matching the daylight and cool screen fill. PHYSICS Real body weight — @ADIL sunk into the sofa; @MOM's natural few-step walk and the settle of her weight as she stops, fabric and towel moving with her. Correct contact shadows under both. No floating props. LIGHTING Warm-neutral daytime daylight raking in from the side window at frame-RIGHT as key — bright with only a gentle warmth. A soft cool screen-light from just behind the lens as fill, washing @ADIL's face in Shot 1 and @MOM's face in Shot 2, brightening on her surprised look. Clear air, minimal haze. WB locked 5200K. COLOR GRADE Neutral daylit room tones as the dominant material, the cool screen glow from behind the lens as the secondary beam on the faces, and a chartreuse #d1ef17 accent living only in a flicker of cool screen light and a glint on a lens flare — color tied to source and surface, not a flat list. AUDIO NO MUSIC. SFX ONLY — diegetic sound and live audio throughout. No score, no soundtrack, no background music of any kind. Low TV soundtrack, soft footsteps as @MOM enters, room tone. Dialogue (English, no subtitles): @MOM (nagging, looking at her son): "What are you watching over there?" — short pause — @MOM (exasperated, fed up): "Come eat — now!" @MOM (surprised, no longer irritated): "Wait… is that 4K?" Then a short half-second of room tone, then end. STYLE 8K photoreal live-action cinematic — practical in-camera realism, naturalistic master cinematography, true material response, fine film grain. OUTPUT SETTINGS 16:9. Real-time throughout. Total duration 11 seconds. POSITIVE LOCKS Exactly one HARD CUT, at 0:07 right after "now"; the camera otherwise holds still in each shot. The camera looks across the room toward the sofa and the plain back wall with its doorway, the side window at the right edge with daylight from the right; the TV @MOM watches is on the wall behind the camera, in her eyeline but out of frame. Shot 1 is a wide low two-shot: @ADIL sunk low-center on the wide sofa, absorbed in the off-frame screen — alive, blinking, breathing, small idle shifts, gaze forward and never into the lens — with @MOM entering through the doorway, taking a few steps and planting herself behind the sofa above her son, looking down as she nags and finishing with "Come eat — now!". Shot 2 is a medium on @MOM at that same spot as her gaze lifts forward to the screen and shifts to genuine surprise — "Wait… is that 4K?" — held for a short beat, moderate and human. @ADIL stays seated and does not rise or leave. @MOM stays in the same spot behind the sofa across the cut — Shot 2 only frames her tighter. Same @ADIL and @MOM identity and wardrobe, same room and towel across the cut. Contact shadows read clearly. Eyes stay natural — no eye glow.

Flawless on the very first try. Watch the continuity on the cut - from the wide shot of the room straight into a tight close-up of her face, same spot, same light. Now, let's stitch all the clips together and see the final result.第一次尝试就完美无缺。看看剪辑的连贯性——从房间的广角镜头直接切到她脸部的紧凑特写,位置一样,光线一样。 现在,让我们把所有片段拼在一起,看看最终结果。

One minute of film. Fantasy, wildlife, a post-apocalyptic chase, a full epic battle - every style, one workflow: assets first, Claude writes the prompt, Seedance 4K renders it.一分钟的电影。奇幻、野生动物、末日追逐、宏大战斗——每种风格,一个流程:先做素材,Claude 写提示词,Seedance 4K 渲染。

The tricks worth keeping from this breakdown: grey backgrounds for every sheet, save everything as elements, the 3/4 angle for locations, switch models when the result looks flat, the red-arrow trick for precise actions, multi-character sheets to kill clone armies, matching duration for video references, and specifying exactly what plays on any screen.这次解析中值得保留的技巧:每张设定表都用灰色背景,把所有东西存为元素,场景用 3/4 角度,结果平淡时就换模型,红箭头技巧用于精确动作,用多角色表来消灭克隆大军,视频参考匹配时长,以及明确指定屏幕上播放的内容。

Want to create your own film in 4K?想制作你自己的 4K 电影吗?

Create your video创建你的视频