wyuzhi /qisi-video-remix

About qisi-video-remix

Give your agent a reference video and this skill writes a new story using the same storytelling trick. You get a full script, image prompts for characters and scenes, a shot list, and video prompts split into 15s or 30s segments. Each segment prompt stands on its own, so you can paste it into any video tool. It only plans and makes no images or video, and the docs are Chinese only.

Best for: Short-video makers who want to turn a format that already works into a fresh story.

• Breaks down how the reference story works, then changes events, choices or the ending
• Writes separate image prompts for each character, each outfit and each scene
• Builds a shot table with timing, camera moves, speakers and sound
• Splits segments at scene, time or outfit changes, not just at the time limit
• Spells out looks, setting and dialogue in every segment prompt, with no "same as above"
• Saves the whole plan as creative-plan.md when the agent can write files

1. 1
   Codex: the command below

   git clone https://github.com/wyuzhi/qisi-video-remix.git "${CODEX_HOME:-$HOME/.codex}/skills/qisi-video-remix"

   Copy
2. 2
   Other agents: put the whole qisi-video-remix folder in the skills directory, with SKILL.md directly inside (drop any -main suffix).
3. 3
   Attach a video, or a synopsis, screenshots or transcript, and send 使用 $qisi-video-remix 改编这个视频。 (remix this video).
4. 4
   Pick a per-segment limit of 15 or 30 seconds; it asks if you don't say.
5. 5
   Ask for edits in plain words, and it updates the affected shots and prompts.