ZenStory AI

Video workflow · 视频工作流

Video Recap Skills

Clip any video into a narrated Chinese recap — scene detection, script, voice-over, cut, subtitles — with a CapCut draft export.

把任意视频剪成中文解说视频——场景检测、写稿、配音、剪辑、字幕——并导出剪映草稿。

505 GitHub stars · 6 skills · MIT license · Format: Video workflow

Source on GitHub English README npx skills add zenstory-ai/video-recap-skills -y -g

What it is

Video Recap Skills is six independent skills plus an orchestrator for Claude Code that turn a source video into a narrated recap. It understands the footage (scene detection, ASR, a vision-language model), writes a director/edit/narration script, generates the voice-over, cuts and assembles with subtitles, and can export a CapCut (剪映) draft for hand-finishing. It runs on ffmpeg and one Xiaomi MiMo API key — no GPU, no model download.

它是什么

Video Recap Skills 是 6 个独立 skill 加一个编排器（Claude Code），把源视频做成解说视频：理解画面（场景检测、ASR、视觉语言模型）、写导演/剪辑/解说稿、生成配音、剪辑与字幕合成，并可导出剪映草稿手工精修。依赖 ffmpeg 和一个小米 MiMo API key——不需要 GPU，也不用下载模型。

Who it is for · 适合谁

  • Recap and commentary channel creators
  • Editors who finish in CapCut
  • Teams turning long footage into short narrated cuts
  • 影视解说 / 说剧类创作者
  • 在剪映里精修的剪辑师
  • 把长素材做成短解说的团队

How it works · 流程

  1. Understand Scene detection + ASR + VLM; optional background research so the model knows who is who.
  2. Script Director, editor and narrator passes write the recap script against the timeline.
  3. Voice Text-to-speech narration.
  4. Cut & assemble Audio mix, subtitles, render; or 'cut first, narrate after' mode.
  5. Export CapCut draft for manual finishing.
  1. 理解 场景检测 + ASR + VLM；可选背景调研，让模型认得人物。
  2. 写稿 导演 / 剪辑 / 解说三道稿子对着时间线写。
  3. 配音 文本转语音。
  4. 剪辑合成 混音、字幕、渲染；或「先剪后配」模式。
  5. 导出 剪映草稿，方便手工精修。

What makes it different · 有什么不同

  • Local pipeline: ffmpeg + one API key, no GPU.
  • CapCut draft export instead of a locked final render.
  • Worked example shipped in the repo.
  • 本地流水线：ffmpeg + 一个 API key，无需 GPU。
  • 导出剪映草稿而不是锁死的成片。
  • 仓库附完整示例。

Terms it uses

解说视频 剪映草稿 先剪后配

Part of ZenStory AI · ZenStory AI 项目

ZenStory AI builds open-source agent skills and workbenches that take a story from research and drafting through to short drama, playable games, and narrated video — running inside Claude Code, Codex, OpenCode, and DeepSeek Harness.

ZenStory AI 打造开源的 AI 故事创作 Agent Skills 与工作台，覆盖从选题、拆解、写作到短剧、可玩游戏与解说视频的全链路，可在 Claude Code、Codex、OpenCode、DeepSeek Harness 等 Agent 环境中运行。

ProjectFormatWhat it does
Oh Story Agent skill pack Turn a coding agent into a complete web-fiction writing workflow — chart scanning, deconstruction, drafting, de-AI editing, covers.
Drama Skills Production skill suite An idea or a novel in; episode scripts, assets, storyboards, image and video prompts, and an independent review out.
Novel to Game Adaptation skill pack Turn a novel in any language into a source-grounded, fully playable game.
Video Recap Skills Video workflow Clip any video into a narrated Chinese recap — scene detection, script, voice-over, cut, subtitles — with a CapCut draft export.
Oh Story DSH DSH plugin The whole story stack inside DeepSeek Harness, with a GUI workbench.
ZenStory Workbench Web workbench Chat to create — an agent-driven novel-writing workbench that keeps files, materials, characters, chapters and versions in one place.