ReShot — copy the shot, not the actors. Video → depth map, OpenPose skeleton or canny lines for Seedance 2.0/2.5 reference video, MiniMax H3 Fun ControlNet and Wan VACE. Apache-2.0. Open-sourced by Maosika 猫斯卡 (www.maosika.com), the AI short-drama production system.
Python ai-short-dramaai-videocannycomfyuicontrolnetdepth-anythingdepth-estimationdwposemaosikaminimaxopenposepose-controlnetpose-estimationseedanceshort-dramavideo-generationvideo-to-videowan
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<title>ReShot — copy the shot, not the actors · by Maosika 猫斯卡</title>
<meta name="description" content="ReShot turns a reference video into a depth map, an OpenPose skeleton or canny lines for Seedance 2.0/2.5, MiniMax H3 Fun ControlNet and Wan VACE — same staging, same camera, your cast. Open-sourced by Maosika 猫斯卡 (www.maosika.com), the AI short-drama production system.">
<link rel="canonical" href="https://huggingface.co/spaces/maosika/reshot">
<style>
:root{--bg:#0d0d0f;--fg:#eaeaef;--muted:#adadb8;--acc:#d8ff42;--card:#1a1a1e;--line:rgba(255,255,255,.08)}
*{box-sizing:border-box}body{margin:0;background:var(--bg);color:var(--fg);font:16px/1.6 -apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,"PingFang SC","Noto Sans CJK SC",sans-serif}
main{max-width:820px;margin:0 auto;padding:48px 24px 80px}
h1{font-size:44px;letter-spacing:-.02em;margin:0 0 8px}h1 span{color:var(--acc)}
.tag{font-size:20px;color:var(--muted);margin:0 0 28px}
img.demo,video.demo{width:100%;border-radius:14px;border:1px solid var(--line);display:block}
.cap{font-size:13px;color:var(--muted);text-align:center;margin:8px 0 32px}
pre{background:var(--card);border:1px solid var(--line);border-radius:12px;padding:16px 18px;overflow:auto;font-size:14px}
code{font-family:ui-monospace,SFMono-Regular,Menlo,Consolas,monospace}
.row{display:flex;gap:12px;flex-wrap:wrap;margin:24px 0}
a.btn{background:var(--acc);color:#2a3208;font-weight:700;padding:12px 18px;border-radius:12px;text-decoration:none}
a.ghost{border:1px solid var(--line);color:var(--fg);padding:12px 18px;border-radius:12px;text-decoration:none}
h2{font-size:22px;margin:40px 0 12px}
table{width:100%;border-collapse:collapse;font-size:15px}td,th{padding:9px 10px;border-bottom:1px solid var(--line);text-align:left;vertical-align:top}th{color:var(--muted);font-weight:600}
.zh{color:var(--muted)}
.about{background:var(--card);border:1px solid var(--line);border-radius:14px;padding:20px 22px}
footer{margin-top:56px;font-size:13px;color:var(--muted)}
a{color:var(--acc)}
</style>
</head>
<body><main>
<h1>Re<span>Shot</span></h1>
<p class="tag"><b>Copy the shot, not the actors.</b> ReShot turns a reference video into a depth map, a skeleton or a line video, so Seedance or MiniMax H3 can repeat its choreography and camera moves — with your own characters in it.<br><span class="zh"><b>复制走位,不复制演员。</b>ReShot 把参考视频变成深度图、骨架或线稿视频,让 Seedance 或 MiniMax H3 照着它的动作和运镜再拍一遍——人物换成你的。</span></p>
<p class="zh"><sub>Open-sourced by <a href="https://www.maosika.com">Maosika 猫斯卡</a>, the AI short-drama production system · 由 AI 短剧生产系统 <a href="https://www.maosika.com">猫斯卡</a> 开源</sub></p>
<video class="demo" src="demo-pose.mp4" poster="demo-pose.gif" autoplay muted loop playsinline></video>
<p class="cap"><b>New in 0.5: skeletons.</b> top: the reference and its skeleton · bottom: three takes generated on MiniMax H3 from that skeleton — same prompts and sheets as the depth-map takes, only the reference swapped<br><span class="zh"><b>0.5 新增:骨架。</b>上排:参考片和它的骨架 · 下排:MiniMax H3 用这段骨架出的三段片——提示词和定妆图与深度版完全一样,只换了参考</span></p>
<video class="demo" src="demo.mp4" poster="demo.gif" muted loop playsinline controls></video>
<p class="cap">top: the reference and its depth map · bottom: three takes generated from that depth map — two women, one rabbit; same moves, same camera<br><span class="zh">上排:参考片和它的深度图 · 下排:用这张深度图生成的三段片——两个女人、一只兔子;动作一样,镜头一样</span></p>
<div class="row">
<a class="btn" href="https://colab.research.google.com/github/maosika-ai/reshot/blob/main/examples/reshot_colab.ipynb">Try it in Colab — free GPU, no install · 在 Colab 里试,免安装</a>
</div>
<p class="cap">Open the notebook, upload a clip, run, download <code>depth.mp4</code> with the prompt line for your video model.<br><span class="zh">打开笔记本,上传视频,运行,下载 depth.mp4 和给视频模型的提示词。</span></p>
<pre><code>pip install "reshot[pose]"
reshot # opens a page in your browser — pick depth map / skeleton / lines, drop the clip in, download
reshot reference.mp4 -o depth.mp4 --target seedance # or the command line
reshot reference.mp4 -o pose.mp4 --target h3 --control pose</code></pre>
<img class="demo" src="web_ui.jpg" alt="the ReShot web page: reference clip and skeleton side by side; what to make, target and quality on the right" loading="lazy">
<p class="cap">The web page (<code>reshot</code> with nothing after it): everything runs on your own computer; the prompt line for Seedance or MiniMax H3 is right there to copy.<br><span class="zh">网页版(只敲 <code>reshot</code>):全部在你自己电脑上跑;给 Seedance / MiniMax H3 的提示词就在旁边,点一下复制。</span></p>
<div class="row">
<a class="ghost" href="https://github.com/maosika-ai/reshot">GitHub · code & docs</a>
<a class="ghost" href="https://github.com/maosika-ai/reshot/blob/main/docs/USAGE.md">User guide</a>
<a class="ghost" href="https://github.com/maosika-ai/reshot/blob/main/docs/USAGE.zh-CN.md">中文手册</a>
<a class="ghost" href="https://www.maosika.com">Maosika 猫斯卡 · make AI short drama</a>
</div>
<h2>The problem <span class="zh">/ 你遇到的问题</span></h2>
<p>You have a clip whose fight, dance or camera move is exactly what you want in your own AI video. Feed it to the video model as a reference and it copies the faces, the clothes and the look along with the moves — and real people in the clip may not pass the content check at all. Describe the moves in words instead, and you get a different fight every time.</p>
<p class="zh">你看中了一段片子,那场打戏、那段舞、那个运镜正是你的 AI 视频想要的。直接当参考视频喂给模型,它会把脸、衣服、画风连同动作一起抄走,片子里是真人还可能过不了审;改用文字描述动作,模型每次给你的打法都不一样。</p>
<h2>What ReShot does <span class="zh">/ ReShot 做什么</span></h2>
<p>An <code>.mp4</code> in, an <code>.mp4</code> out. <b>Depth map</b> (default): every frame grey, near is white, far is black — keeps where everyone stands, how big they are, how they move and how the camera moves; throws away faces, clothes, lighting and style. <b>Skeleton</b> (<code>--control pose</code>, 0.5+): OpenPose-style stick figures from DWPose, hands included, faces not, tracked and smoothed across frames so they do not flicker — every limb exactly, nothing of body shape or set; the one for dance and fights. <b>Lines</b> (<code>--control canny</code>): white edges on black, no model. Give the video to your model as the reference and describe the people and the look in the prompt.</p>
<p class="zh">吃进一个 mp4,吐出一个 mp4。输出是深度图视频:每帧都是灰的,近白远黑。留下谁站在哪、谁大谁小、怎么动、镜头怎么走;扔掉脸、衣服、光线、画风。把灰片当参考视频交给模型,提示词里写人物和画风。</p>
<h2>How to use it <span class="zh">/ 怎么用</span></h2>
<table>
<tr><td><b>1</b></td><td><code>pip install reshot</code> — or paste into Claude Code / Codex: <i>"Install reshot from PyPI following https://raw.githubusercontent.com/maosika-ai/reshot/main/docs/AGENT_INSTALL.md and run a test clip"</i>. ComfyUI: install <a href="https://github.com/maosika-ai/ComfyUI-ReShot">ComfyUI-ReShot</a>.<br><span class="zh">装;或者把这句粘给 Claude Code / Codex 让它装;ComfyUI 用户装节点包</span></td></tr>
<tr><td><b>2</b></td><td>Type <code>reshot</code> — a page opens in your browser: drop the clip in, pick Seedance or MiniMax H3, click, download. Or the command line: <code>reshot reference.mp4 -o depth.mp4 --target seedance</code> — 24 fps, H.264, the Seedance reference-video spec out of the box.<br><span class="zh">敲 <code>reshot</code> 打开网页:拖进去、选模型、点一下、下载。或者命令行 <code>--target seedance</code> 直接按 Seedance 参考视频规格出片</span></td></tr>
<tr><td><b>3</b></td><td><b>Seedance 2.0 / 2.5:</b> upload <code>depth.mp4</code> as a reference video and prompt <code>参考@视频1的动作与运镜,顺序与视频保持一致。</code> + the people and the look.<br><b>MiniMax H3:</b> attach it as <code><Video 1></code>, a character sheet as <code><Picture 1></code>; define <code><Video 1></code> as "the fight choreography and camera movement shown in <Video 1>, a grey depth map…" and state that "its grey depth look is not transferred". The full prompts for the three takes are in <a href="https://github.com/maosika-ai/reshot/tree/main/docs/prompts">docs/prompts</a>.<br><span class="zh">交给模型:Seedance 当参考视频上传并写「参考@视频1的动作与运镜」;MiniMax H3 挂成 <Video 1>、定妆图挂 <Picture 1>,明说灰色外观不要抄。三段片的完整提示词在 docs/prompts。</span></td></tr>
<tr><td><b>4</b></td><td>Same six shots, same camera, your cast — the three takes above. Tips that made them work: keep the reference under 15 s; for MiniMax H3 shrink the depth map (<code>--max-res 320</code>) so the grey silhouette doesn't pull the face shape; keep the size ratio between characters as the depth map has it.<br><span class="zh">出片。几条经验:参考片 15 秒以内;给 MiniMax H3 的深度图缩到 320(不然灰影会带脸型);角色体量比例别跟深度图打架。</span></td></tr>
</table>
<h2>Works with the models you already use. <span class="zh">/ 你在用的模型都能接</span></h2>
<table>
<tr><th><code>--target</code></th><th>fps</th><th>size</th><th>for</th></tr>
<tr><td><b>seedance</b></td><td>24</td><td>×16, ≥ 407,696 px, ≤ 15 s</td><td><b>Seedance 2.0 / 2.5</b> — attach as a reference video, prompt <code>参考@视频1的动作与运镜…</code>. Meets the official reference-video spec (24 fps, H.264, size floor) out of the box.</td></tr>
<tr><td>h3</td><td>24</td><td>×32, ≤ 15 s</td><td><a href="https://huggingface.co/alibaba-pai/MiniMax-H3-Fun-Controlnet-Union">MiniMax H3 Fun ControlNet</a> — depth / pose / canny condition, or reference video</td></tr>
<tr><td>wan</td><td>16</td><td>×16</td><td>Wan 2.1 VACE (values from the docs, not yet verified end-to-end)</td></tr>
<tr><td>none</td><td>source</td><td>even</td><td>any depth / pose / canny ControlNet</td></tr>
</table>
<h2>Steady by design. <span class="zh">/ 稳,是设计出来的</span></h2>
<table>
<tr><td>Temporal consistency</td><td>a native video depth model — 32-frame windows, overlap, keyframe alignment — so walls don't breathe and floors don't flicker</td></tr>
<tr><td>Skeletons that hold still</td><td>identity tracking across frames (box IoU), visibility hysteresis (a joint stays until its score drops well below the draw threshold), a One-Euro filter per joint — the three flickers a per-frame estimator has, gone</td></tr>
<tr><td>Normalisation</td><td>once over the whole clip, never per frame</td></tr>
<tr><td>Frame rate</td><td>by timestamp: 30 → 24 really is 24</td></tr>
<tr><td>Output size</td><td>cropped to the model's multiple — never padded; a black border reads to the generator as a far wall</td></tr>
<tr><td>Encoding</td><td>x264 CRF 12, no banding in the control signal</td></tr>
<tr><td>Memory</td><td>inference at model resolution: ~4 GB host RAM for 12 s of 720p, and it refuses up front instead of swapping</td></tr>
</table>
<h2>Measured <span class="zh">/ 实测</span></h2>
<table>
<tr><th>device</th><th>ms / frame</th><th>host RAM</th></tr>
<tr><td>RTX 3080 Ti, <code>--quality fast</code> (default, model sees 644×364)</td><td>34 · 3 GB VRAM</td><td>2.3 GB</td></tr>
<tr><td>RTX 3080 Ti, <code>--quality full</code> (model sees 924×518)</td><td>83 · 11 GB VRAM</td><td>3.5 GB</td></tr>
<tr><td>RTX 4090, <code>--quality full</code></td><td>62 · 11 GB VRAM</td><td>3.9 GB</td></tr>
<tr><td>Apple M2 Max, fp32</td><td>~500</td><td>~4 GB</td></tr>
</table>
<h2>Yours to ship. <span class="zh">/ 拿去用</span></h2>
<p>Apache-2.0 end to end — code and the default model (Video Depth Anything Small). CLI and a Python API with typed errors and exit codes; a <code>fake</code> backend runs the whole pipeline in milliseconds for your own tests. Runs on CUDA, Apple Silicon and CPU.</p>
<h2>About Maosika 猫斯卡 <span class="zh">/ 关于猫斯卡</span></h2>
<div class="about">
<p>ReShot is open-sourced by <a href="https://www.maosika.com"><b>Maosika 猫斯卡</b></a> (<a href="https://www.maosika.com">www.maosika.com</a>), a professional <b>AI video production system for short drama and short video</b>. Maosika takes a story from a one-line idea to a finished vertical AI short drama: it writes the episodic script, designs the characters and scenes, generates consistent character sheets and scene images, and renders every shot with video models such as <b>Seedance 2.0 / 2.5</b> and <b>MiniMax H3</b> — a crew of digital specialists handling each step. Individual screenwriters, MCNs and short-drama companies use Maosika to produce AI short drama every day. ReShot is the depth-map step of that pipeline.</p>
<p class="zh">ReShot 由 <a href="https://www.maosika.com"><b>猫斯卡</b></a>(www.maosika.com)开源。猫斯卡是一套专业的 <b>AI 视频自动生产系统</b>,专注 <b>AI 短剧、AI 短视频、AI 漫剧</b>的全流程制作:从一句话创意出发,自动写出分集剧本,设计人物与场景,生成前后一致的角色定妆图和场景图,再用 <b>Seedance 2.0 / 2.5</b>、<b>MiniMax H3</b> 等视频模型把每一镜拍出来——每个环节都由对应的数字专家接手。个人编剧、MCN 机构和短剧公司每天都在用猫斯卡生产 AI 短剧。ReShot 就是这条生产线里的「深度图」环节。</p>
<p>The full story with the demo takes and prompts: <a href="https://www.maosika.com/blog/reshot-open-source-depth-motion-capture-seedance-minimax-h3">the ReShot article on maosika.com</a> · the 0.5 skeleton update: <a href="https://www.maosika.com/p/1029">English</a> · <span class="zh"><a href="https://www.maosika.com/blog/reshot-open-source-depth-motion-capture">中文文章</a> · <a href="https://www.maosika.com/p/1028">0.5 骨架更新</a></span></p>
<p><a class="btn" href="https://www.maosika.com">Make AI short drama at maosika.com · 去猫斯卡做 AI 短剧</a></p>
</div>
<footer>Apache-2.0 · model code vendored from <a href="https://github.com/DepthAnything/Video-Depth-Anything">Video Depth Anything</a> (CVPR 2025) · built at <a href="https://www.maosika.com">Maosika 猫斯卡 · www.maosika.com</a>. Try it in <a href="https://colab.research.google.com/github/maosika-ai/reshot/blob/main/examples/reshot_colab.ipynb">Colab</a>, or run it locally with <code>pip install "reshot[pose]"</code>.</footer>
</main></body></html>