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
docs: skeleton demo — the three takes re-made on MiniMax H3 from the pose video (demo-pose.gif/mp4), prompts with seeds; replaces the three-panel GIF
wangfeng8f2a2e8ec7c0
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 723cea9..22260df 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -23,6 +23,9 @@- Colab notebook and `docs/AGENT_INSTALL.md` cover the new types.- Pose and canny are read at the output size (no "model resolution"), with their own, muchsmaller host-RAM estimate.+- Verified on MiniMax H3: the three demo takes re-made with the skeleton as `<Video 1>` (same+ prompts and sheets) reproduce the six beats like the depth takes — `docs/demo-pose.gif`,+ prompts with seeds in `docs/prompts/*_pose.txt`.- Verified on MiniMax H3: the three demo takes re-made with the skeleton as `<Video 1>` (sameprompts and sheets) reproduce the six beats like the depth takes; prompts + seeds in`docs/prompts/*_pose.txt`, comparison in `docs/img/pose_vs_depth_takes.jpg`.diff --git a/README.md b/README.mdindex f379375..0af510c 100644--- a/README.md+++ b/README.md@@ -32,8 +32,8 @@ ReShot takes an `.mp4` in and writes an `.mp4` out. The output is a control vide| **`canny`** | white edge lines on black | composition and outlines | colour and shading — but *not* the outline of a face or a costume | when you want the picture's layout, not only the people |<p align="center"><img src="docs/img/step1_reference.jpg" width="300" alt="reference frame"> <img src="docs/img/step2_depth.jpg" width="300" alt="the same frame as a depth map"> <img src="docs/img/step2_pose.jpg" width="300" alt="the same frame as a skeleton"></p>-<p align="center"><img src="docs/demo-controls.gif" width="864" alt="the corridor fight as reference, depth map and skeleton, playing side by side"></p>-<p align="center"><sub>Reference · depth map · skeleton, the same 12 seconds. Four stills at full size: <a href="docs/img/controls_strip.jpg">controls_strip.jpg</a>.</sub></p>+<p align="center"><img src="docs/demo-pose.gif" width="720" alt="the fight scene, its skeleton, and three new takes generated from the skeleton"></p>+<p align="center"><sub>The same demo made from the <b>skeleton</b> instead of the depth map. Top: the reference and its skeleton video. Bottom: three takes generated on MiniMax H3 from that skeleton — same prompts and character sheets as the depth takes, only <code><Video 1></code> swapped. <a href="docs/demo-pose.mp4">Full-resolution clip</a> · prompts with seeds in <a href="docs/prompts/">docs/prompts/</a> (<code>*_pose.txt</code>). Four stills of reference · depth · skeleton: <a href="docs/img/controls_strip.jpg">controls_strip.jpg</a>.</sub></p>You give that video to your video model as the reference and describe the people and the look in the prompt. The model takes the moves from the video and everything else from your words.@@ -134,7 +134,7 @@ Left to right: [Jiang Xue](docs/prompts/take1_jiangxue_armor.txt) in bronze armo**Doesn't:** faces (use a character sheet), clothes, lighting, colour, props in detail, and anything smaller than a hand. Those come from your prompt and your reference images.-**Depth or pose?** Depth carries the whole picture — where the walls are, who is bigger, a pipe being torn off — and works on anything, people or not. Pose carries only the people, but every limb and finger exactly, and nothing of their body shape or the set. Use pose for dance and martial arts, depth for everything else; when in doubt make both (`--control depth,pose`) and see which one the model follows better.+**Depth or pose?** Depth carries the whole picture — where the walls are, who is bigger, a pipe being torn off — and works on anything, people or not. Pose carries only the people, but every limb and finger exactly, and nothing of their body shape or the set. Use pose for dance and martial arts, depth for everything else; when in doubt make both (`--control depth,pose`) and see which one the model follows better. We re-made the three demo takes on MiniMax H3 with the skeleton as `<Video 1>` (same prompts, same sheets): all three reproduce the six beats like the depth takes, and the throw reads more clearly from the skeleton — see the second GIF above. One take per condition with different seeds, so read it as "pose works as a reference video too", not "pose is better".We tested that on MiniMax H3 with the three demo takes: same prompts, same character sheets, only `<Video 1>` swapped from the depth map to the skeleton video (and the two sentences that describe it). All three skeleton takes reproduce the six beats — close-up, brawl, the throw, the panel, the walk out past the fallen guards — at the same level as the depth takes; the throw reads more clearly from the skeleton. One take per condition, different seeds, so read it as "pose works as a reference video too", not as "pose is better". The prompts with seeds are in [`docs/prompts/`](docs/prompts/) (`*_pose.txt`).diff --git a/README.zh-CN.md b/README.zh-CN.mdindex 1f9d978..418ffdf 100644--- a/README.zh-CN.md+++ b/README.zh-CN.md@@ -32,8 +32,8 @@ ReShot 吃进一个 `.mp4`,吐出一个 `.mp4`。输出是控制视频,三| **`canny`** | 黑底白线的边缘图 | 构图和轮廓 | 颜色和明暗——但脸型和服装的轮廓**留着** | 想要整个画面的布局、不只是人的时候 |<p align="center"><img src="docs/img/step1_reference.jpg" width="300" alt="参考片的一帧"> <img src="docs/img/step2_depth.jpg" width="300" alt="同一帧的深度图"> <img src="docs/img/step2_pose.jpg" width="300" alt="同一帧的骨架"></p>-<p align="center"><img src="docs/demo-controls.gif" width="864" alt="走廊武打的参考片、深度图和骨架并排播放"></p>-<p align="center"><sub>参考片 · 深度图 · 骨架,同一段 12 秒。四张全尺寸静帧:<a href="docs/img/controls_strip.jpg">controls_strip.jpg</a>。</sub></p>+<p align="center"><img src="docs/demo-pose.gif" width="720" alt="武打参考片、它的骨架、用骨架生成的三段新片"></p>+<p align="center"><sub>同一个演示,参考换成<b>骨架</b>。上排:参考片和它的骨架视频。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 <code><Video 1></code>。<a href="docs/demo-pose.mp4">高清原片</a> · 带 seed 的提示词在 <a href="docs/prompts/">docs/prompts/</a>(<code>*_pose.txt</code>)。参考 · 深度 · 骨架四张静帧:<a href="docs/img/controls_strip.jpg">controls_strip.jpg</a>。</sub></p>你把这段视频当参考交给视频模型,提示词里写人物和画风。动作和镜头模型从片里读,其余全听你的。@@ -134,7 +134,7 @@ is transferred onto <Subject 1> and <Subject 2>; its grey depth look is not tran**不能:** 脸(用定妆图)、衣服、光线、颜色、道具细节,以及比手还小的东西。这些全靠你的提示词和参考图。-**深度图还是骨架?** 深度图带走整个画面——墙在哪、谁更大、一根管子被扯下来——人和非人都管用。骨架只带人,但每根肢体每根手指都准,体型和场景一点不带。跳舞、武打选骨架,其余选深度图;拿不准就两个都出(`--control depth,pose`),看模型跟哪个跟得更好。+**深度图还是骨架?** 深度图带走整个画面——墙在哪、谁更大、一根管子被扯下来——人和非人都管用。骨架只带人,但每根肢体每根手指都准,体型和场景一点不带。跳舞、武打选骨架,其余选深度图;拿不准就两个都出(`--control depth,pose`),看模型跟哪个跟得更好。我们在 MiniMax H3 上把演示的三段片用骨架当 `<Video 1>` 重出了一遍(提示词、定妆图完全一样):三条都复现了六个节拍,和深度版一个水平,过肩摔那一镜骨架版更清楚——见上面第二张动图。每种条件只出了一条、seed 不同,所以只能读成「骨架当参考视频也行」,不是「骨架更好」。我们在 MiniMax H3 上用演示的三段片做了实测:提示词、定妆图完全一样,只把 `<Video 1>` 从深度图换成骨架视频(连带改了描述它的两句话)。三条骨架版都复现了六个节拍——特写、群殴、过肩摔、撞控制台、跨过倒地守卫走出门——和深度版一个水平,过肩摔那一镜骨架版更清楚。每种条件只出了一条、seed 不同,所以只能读成「骨架当参考视频也行」,不能读成「骨架更好」。带 seed 的提示词在 [`docs/prompts/`](docs/prompts/)(`*_pose.txt`)。diff --git a/docs/demo-controls.gif b/docs/demo-controls.gifdeleted file mode 100644index 4a8f8c5..0000000Binary files a/docs/demo-controls.gif and /dev/null differdiff --git a/docs/demo-pose.gif b/docs/demo-pose.gifnew file mode 100644index 0000000..88d264aBinary files /dev/null and b/docs/demo-pose.gif differdiff --git a/docs/demo-pose.mp4 b/docs/demo-pose.mp4new file mode 100644index 0000000..9cbe313Binary files /dev/null and b/docs/demo-pose.mp4 differdiff --git a/docs/img/pose_vs_depth_takes.jpg b/docs/img/pose_vs_depth_takes.jpgdeleted file mode 100644index 196d55a..0000000Binary files a/docs/img/pose_vs_depth_takes.jpg and /dev/null differdiff --git a/docs/prompts/take1_jiangxue_armor_pose.txt b/docs/prompts/take1_jiangxue_armor_pose.txtindex 81c5586..08508be 100644--- a/docs/prompts/take1_jiangxue_armor_pose.txt+++ b/docs/prompts/take1_jiangxue_armor_pose.txt@@ -1,5 +1,4 @@-# MiniMax H3, <Video 1> = the OpenPose skeleton video (reshot --control pose --target h3, scaled to 320x176), seed 1093962678, 12 s 16:9 720p, 340 s on one H3 machine (2026-09-14).-# Identical to the depth-map take except the two sentences that define <Video 1> and say its look is not to be copied.+# MiniMax H3 · <Video 1> = the skeleton video (reshot --control pose --target h3, scaled to 320x176) · seed 2007422674 · 330 ssubject_definitions:<Subject 1> is the young woman shown in <Picture 1> and in close-up in <Picture 2>, Jiang Xue, a warrior with a high ponytail, red flame tattoos across her chest and shoulders and a dark leather-and-bronze armored corset; both pictures are used only for her appearance and her face must match <Picture 2> exactly.diff --git a/docs/prompts/take2_suwan_qipao_pose.txt b/docs/prompts/take2_suwan_qipao_pose.txtindex 56512cd..c59333f 100644--- a/docs/prompts/take2_suwan_qipao_pose.txt+++ b/docs/prompts/take2_suwan_qipao_pose.txt@@ -1,5 +1,4 @@-# MiniMax H3, <Video 1> = the OpenPose skeleton video (reshot --control pose --target h3, scaled to 320x176), seed 1488156354, 12 s 16:9 720p, 340 s on one H3 machine (2026-09-14).-# Identical to the depth-map take except the two sentences that define <Video 1> and say its look is not to be copied.+# MiniMax H3 · <Video 1> = the skeleton video (reshot --control pose --target h3, scaled to 320x176) · seed 327102747 · 334 ssubject_definitions:<Subject 1> is the young woman shown in <Picture 1> and in close-up in <Picture 2>, Su Wan, an elegant beauty with black hair in a low side bun, a pearl earring and a dark green velvet qipao; both pictures are used only for her appearance and her face must match <Picture 2> exactly.diff --git a/docs/prompts/take3_rabbit_boxer_pose.txt b/docs/prompts/take3_rabbit_boxer_pose.txtindex 2c28e60..b5821c9 100644--- a/docs/prompts/take3_rabbit_boxer_pose.txt+++ b/docs/prompts/take3_rabbit_boxer_pose.txt@@ -1,5 +1,4 @@-# MiniMax H3, <Video 1> = the OpenPose skeleton video (reshot --control pose --target h3, scaled to 320x176), seed 927905444, 12 s 16:9 720p, 328 s on one H3 machine (2026-09-14).-# Identical to the depth-map take except the two sentences that define <Video 1> and say its look is not to be copied.+# MiniMax H3 · <Video 1> = the skeleton video (reshot --control pose --target h3, scaled to 320x176) · seed 927905444 · 328 ssubject_definitions:<Subject 1> is the anthropomorphic rabbit shown in <Picture 1>, a female rabbit with cream-tan fur, dark-tipped upright ears, big green eyes, a tiny silver ear stud, a grey hoodie, black shorts and white hand wraps, standing upright like a person, about 1.5 meters tall (baseline 1x), her head reaching the wolf's shoulder; the picture (face on the left, full body on the right) is used only for her appearance and her face, fur color and outfit must match it exactly. She is the only rabbit in the entire film: there is exactly one rabbit, and every other character is a predator of <Subject 2>.