3D White-Model Motion Transfer Workflow
Using Seedance 2.0 fast to convert a reference video into a 3D white model, then rendering new footage with 2.5 — the whole action and camera movement gets carried over. The value here is the workflow itself, and both prompt stages are preserved in full.
PROMPT · 3D White-Model Conversion Prompt
Create with AI3D White-Model Conversion Prompt：
将参考视频转换为极简单色 CG 白模视频风格：。严格逐帧保留参考视频的运镜、构图、镜头节奏、人物动作、姿态及手中物体的运动；将人物主体完全替换为 image。
人物是平滑、无发、无表情的白色人体模型，头部、皮肤、上衣、裤子和手持物统一采用无纹理的浅灰白哑光材质。保留简洁、真实的三维体积、衣物轮廓和手指姿势，但不要皮肤毛孔、布料纹理、图案、logo、饰品或材质反光。使用柔和的正前上方棚拍光，使额头、肩部、胸口和手部呈亮白，侧面、手臂内侧和衣物凹陷处自然过渡为浅灰至中灰；阴影柔和克制，带轻微环境遮蔽。
为将原场景（背景）转化为纯黑空间中的抽象灰阶空间层级：人物脚下保留一个简洁、无材质的中灰色承托平面；根据原视频中镜头和地形的高低关系，在前景、中景与远景形成若干平滑、低细节的灰色高度平面或柔和阶层。近处与较高位置略亮，远处、较低位置逐渐压暗至深灰或黑色，以清楚表现镜头升降、环绕和空间纵深。。仅允许黑、白、灰，不得出现任何其他颜色。不要深度摄像头效果、不要距离图、不要扫描线、不要网格、不要线框、不要科技 HUD、不要真实摄影质感、不要文字、水印、logo 或 UI。不要声音、音效和背景音乐。
轉3D白模提示詞：
將參考影片轉換為極簡單色 CG 白模影片風格：。嚴格逐幀保留參考影片的運鏡、構圖、鏡頭節奏、人物動作、姿態及手中物體的運動；將人物主體完全替換為 image。
人物是平滑、無發、無表情的白色人體模型，頭部、皮膚、上衣、褲子和手持物統一採用無紋理的淺灰白啞光材質。保留簡潔、真實的三維體積、衣物輪廓和手指姿勢，但不要皮膚毛孔、布料紋理、圖案、logo、飾品或材質反光。使用柔和的正前上方棚拍光，使額頭、肩部、胸口和手部呈亮白，側面、手臂內側和衣物凹陷處自然過渡為淺灰至中灰；陰影柔和剋制，帶輕微環境遮蔽。
為將原場景（背景）轉化為純黑空間中的抽象灰階空間層級：人物腳下保留一個簡潔、無材質的中灰色承託平面；根據原影片中鏡頭和地形的高低關係，在前景、中景與遠景形成若干平滑、低細節的灰色高度平面或柔和階層。近處與較高位置略亮，遠處、較低位置逐漸壓暗至深灰或黑色，以清楚表現鏡頭升降、環繞和空間縱深。。僅允許黑、白、灰，不得出現任何其他顏色。不要深度攝像頭效果、不要距離圖、不要掃描線、不要網格、不要線框、不要科技 HUD、不要真實攝影質感、不要文字、水印、logo 或 UI。不要聲音、音效和背景音樂。
3D white-model conversion prompt:
Convert the reference video into a minimalist, monochrome CG white-model video style. Strictly preserve, frame by frame, the reference video's camera movement, composition, shot pacing, character motion, poses, and the movement of any object in hand; completely replace the human subject with [image].
The character is a smooth, hairless, expressionless white mannequin figure; the head, skin, top, pants, and any held object are all rendered in a uniform, textureless light gray-white matte material. Preserve clean, realistic three-dimensional volume, clothing silhouette, and finger poses, but do not include skin pores, fabric texture, patterns, logos, accessories, or material reflections. Use soft, front-top studio lighting so that the forehead, shoulders, chest, and hands appear bright white, while the sides, inner arms, and clothing recesses transition naturally from light gray to mid-gray; shadows should be soft and restrained, with a slight ambient occlusion.
Transform the original scene (background) into an abstract grayscale spatial hierarchy set within a pure black space: keep a simple, textureless mid-gray support plane beneath the character's feet; based on the relative height relationships of the camera and terrain in the original video, form several smooth, low-detail gray height planes or soft tiers across the foreground, midground, and background. Areas that are closer or higher should be slightly brighter, while areas that are farther or lower should gradually darken to dark gray or black, clearly conveying camera rise/fall, orbiting movement, and spatial depth. Only black, white, and gray are allowed -- no other colors may appear. No depth-camera effect, no distance map, no scan lines, no grid, no wireframe, no sci-fi HUD, no realistic photographic look, no text, watermark, logo, or UI. No sound, sound effects, or background music.
3D白モデル変換プロンプト：
参考動画を極めてミニマルな単色のCG白モデル映像スタイルへと変換する：参考動画のカメラワーク、構図、ショットのリズム、人物の動作、姿勢、手中の物体の動きを厳密にフレーム単位で保持すること。人物の主体を完全にimageへと置き換える。
人物は滑らかで、髪もなく、表情もない白い人体モデルであり、頭部、肌、上着、パンツ、手持ちの物は統一して無地無質感の淡いグレーホワイトのマット素材とする。シンプルでリアルな三次元の立体感、衣服の輪郭、指の姿勢は保つが、肌の毛穴、布地の質感、模様、ロゴ、装飾品、素材の反射は入れない。柔らかな真正面上方からのスタジオライトを使用し、額、肩、胸、手を明るい白に、側面、腕の内側、衣服のくぼんだ部分は自然に淡いグレーから中間グレーへと移行させる。影は柔らかく抑制的で、わずかな環境遮蔽（アンビエントオクルージョン）を伴う。
元の場面（背景）を、純黒の空間の中の抽象的なグレースケールの空間の階層へと変換する：人物の足元にはシンプルで無質感の中間グレーの支持面を残す。元の動画におけるカメラと地形の高低関係に基づき、前景、中景、遠景にいくつかの滑らかで低ディテールの灰色の高度平面あるいは柔らかな段差を形成する。近くと高い位置はやや明るく、遠くと低い位置は次第に暗い灰色または黒へと落ち込ませ、カメラの上昇・下降、周回、空間の奥行きを明瞭に表現する。黒、白、グレーのみを許容し、その他の色は一切現れてはならない。デプスカメラのエフェクトなし、距離マップなし、走査線なし、グリッドなし、ワイヤーフレームなし、テクノロジー風HUDなし、リアルな撮影の質感なし、テキストなし、透かしなし、ロゴなし、UIなし。音声、効果音、背景音楽はなし。
3D 화이트모델 변환 프롬프트:
참조 영상을 미니멀한 단색 CG 화이트모델 영상 스타일로 변환. 참조 영상의 카메라 워크, 구도, 샷 리듬, 인물 동작, 포즈 및 손에 든 사물의 움직임을 프레임 단위로 엄격히 유지; 인물 주체를 image로 완전히 교체.
인물은 매끄럽고 머리카락이 없으며 무표정한 흰색 인체 모형, 머리·피부·상의·바지·손에 든 물건 모두 질감 없는 연한 회백색 무광 재질로 통일. 간결하고 사실적인 3D 볼륨감, 의상 실루엣, 손가락 포즈는 유지하되 피부 모공, 원단 질감, 패턴, 로고, 장신구, 재질 반사는 없음. 부드러운 정면 상단 스튜디오 조명을 사용해 이마·어깨·가슴·손 부위가 밝은 흰색을 띠게 하고, 측면·팔 안쪽·의상 오목한 부분은 연한 회색에서 중간 회색으로 자연스럽게 전환; 그림자는 부드럽고 절제되며 옅은 앰비언트 오클루전 포함.
원본 장면(배경)을 순수 검정 공간 속 추상적 그레이스케일 공간 계층으로 전환: 인물의 발 아래에는 간결하고 질감 없는 중간 회색 지지면을 유지; 원본 영상의 카메라와 지형의 고저 관계에 따라 전경·중경·원경에 매끄럽고 디테일이 낮은 회색 높이 평면 또는 부드러운 단차를 여러 개 형성. 가까운 곳과 높은 위치는 약간 더 밝게, 먼 곳과 낮은 위치는 점차 짙은 회색 또는 검정으로 어두워져 카메라의 상승·하강, 회전, 공간 깊이감을 명확히 표현. 흑·백·회색만 허용, 그 외의 어떤 색상도 등장 금지. 뎁스 카메라 효과 없음, 거리 맵 없음, 스캔라인 없음, 그리드 없음, 와이어프레임 없음, SF HUD 없음, 실사 촬영 질감 없음, 텍스트·워터마크·로고·UI 없음. 소리, 음향 효과, 배경음악 없음.
✍️ Editor’s Notes
This is a textbook “negative-word flood” color-control prompt — it just keeps hammering the model's default tendency toward rich textures, reflections, and color with wave after wave of “don't,” leaving nothing but three tones of black/white/gray plus a single ground plane to sell spatial depth. For heavily-constrained stylization tasks like converting to a white-model render, positive description usually struggles to nail down exactly what you want, so stacking negative words to draw a hard line around “what must never show up” turns out to be more stable — and this one pushes that approach to the absolute limit.
Workflow Prompt
PROMPT · Full Prompt
Create with AI如何通过3D白模 视频 实现 动作+运镜 迁移：
Seedance2.0 fast 完成 3D白模 转绘 （省级分）
Seedance 2.5 完成新画面的渲染 （效果 最好，最稳定）
具体流程：
1.上传参考视频，选择Seedance2.0 fast ，480P，输入转3D白模的提示词：（见评论区）
2.将3D白模视频+ 角色图， 选择 Seedance 2.5 ，对画质要求高的选720P，要求不高的选480P，输入新的Scene Prompt：
将参考视频中视频 - 2026-08-19T180755.671的主体人物 换成Image 9 (2) 。生成一段 该女主在末世现代都市的顶楼奔跑的画面 ，镜头跟随女主持续往前奔跑，随后 跳起抓住直升机坠下的绳子。随后，镜头跟随下摇，展示女主底下，正在燃烧爆炸的城市和街道。完全真人写实质感，电影感，仅生成音效，不要生成背景音乐。
如何透過3D白模 影片 實現 動作+運鏡 遷移：
Seedance2.0 fast 完成 3D白模 轉繪 （省級分）
Seedance 2.5 完成新畫面的渲染 （效果 最好，最穩定）
具體流程：
1.上傳參考影片，選擇Seedance2.0 fast ，480P，輸入轉3D白模的提示詞：（見評論區）
2.將3D白模影片+ 角色圖， 選擇 Seedance 2.5 ，對畫質要求高的選720P，要求不高的選480P，輸入新的場景提示詞：
將參考影片中影片 - 2026-08-19T180755.671的主體人物 換成Image 9 (2) 。生成一段 該女主在末世現代都市的頂樓奔跑的畫面 ，鏡頭跟隨女主持續往前奔跑，隨後 跳起抓住直升機墜下的繩子。隨後，鏡頭跟隨下搖，展示女主底下，正在燃燒爆炸的城市和街道。完全真人寫實質感，電影感，僅生成音效，不要生成背景音樂。
How to achieve motion + camera movement transfer via 3D white-model video:
Seedance 2.0 fast completes the 3D white-model conversion (saves credits)
Seedance 2.5 completes the rendering of the new footage (best and most stable results)
Specific workflow:
1. Upload the reference video, select Seedance 2.0 fast, 480P, and enter the prompt for converting to a 3D white model (see the comments section):
2. Take the 3D white-model video + the character image, select Seedance 2.5 -- choose 720P if you need high image quality, or 480P if quality requirements are lower -- and enter the new scene prompt:
Replace the main subject in the reference video "视频 - 2026-08-19T180755.671" with Image 9 (2). Generate a scene of this female lead running across a rooftop in a post-apocalyptic modern city, with the camera following her as she keeps running forward, then jumping up to grab a rope dropped from a helicopter. After that, the camera tilts down following her, revealing the burning, exploding city and streets below her. Fully photorealistic human quality, cinematic feel, generate sound effects only, do not generate background music.
3D白モデル動画を通じて動作＋カメラワークの転写を実現する方法：
Seedance2.0 fastで3D白モデルへの変換を行う（コマ数を節約）
Seedance 2.5で新しい画面のレンダリングを行う（効果は最良で最も安定）
具体的な流れ：
1. 参考動画をアップロードし、Seedance2.0 fast、480Pを選択し、3D白モデルへの変換プロンプトを入力する：（コメント欄参照）
2. 3D白モデル動画＋キャラクター画像を、Seedance 2.5を選択し、画質を求める場合は720Pを、そうでなければ480Pを選択し、新しい場面のプロンプトを入力する：
参考動画である動画 - 2026-08-19T180755.671の主体人物をImage 9 (2)に差し替える。終末後の現代都市の屋上を走る女性主人公の映像を生成する。カメラは女性主人公が前方へ走り続けるのに合わせて追従し、その後彼女はジャンプしてヘリコプターから垂れ下がるロープを掴む。続いて、カメラは追従しながら下方へパンし、女性主人公の下方で燃え盛り爆発している都市と通りを映し出す。完全に実写のリアルな質感、映画的な雰囲気、効果音のみを生成し、背景音楽は生成しないこと。
3D 화이트모델 영상으로 동작+카메라 워크 전이를 구현하는 방법:
Seedance 2.0 fast로 3D 화이트모델 변환 완료 (크레딧 절약)
Seedance 2.5로 새 화면 렌더링 완료 (효과 최고, 가장 안정적)
구체적인 절차:
1. 참조 영상을 업로드, Seedance 2.0 fast 선택, 480P, 3D 화이트모델 변환 프롬프트 입력: (댓글란 참고)
2. 3D 화이트모델 영상 + 캐릭터 이미지를 준비, Seedance 2.5 선택, 화질 요구가 높으면 720P, 낮으면 480P 선택, 새로운 장면 프롬프트 입력:
참조 영상 중 '视频 - 2026-08-19T180755.671'의 주체 인물을 Image 9 (2)로 교체. 이 여주인공이 종말 이후 현대 도시의 옥상에서 달리는 장면을 생성, 카메라는 여주인공을 따라 계속 앞으로 달리는 모습을 추적한 뒤 그녀가 뛰어올라 헬기에서 내려온 로프를 붙잡음. 이어서 카메라는 아래로 틸다운하며 그녀 아래에서 불타고 폭발하는 도시와 거리를 보여줌. 완전한 실사 인물 질감, 시네마틱한 느낌, 음향 효과만 생성하고 배경음악은 생성하지 않음.
✍️ Editor’s Notes
This one isn't a visual prompt at all, it's a workflow note, and what it reveals is a relay technique the community has already dialed in: “run the 2.0 fast tier for the white-model conversion to save compute, then hand off to 2.5 for the final render.” Get the motion skeleton running fast on the cheap tier first, then bring in the flagship model for the final image. That division-of-labor thinking is honestly more valuable as a reference than the prompt text itself — it shows players are already treating different model tiers as separate stations on an assembly line instead of grinding one single model to death trying to make it handle the whole workload.
Related Works
׋![]()
›