Claude Real Video
2,191 starsLets an LLM actually watch a video: since models cannot ingest video directly, the skill extracts scene-aware, deduplicated keyframes plus a transcript first, then reads those to summarize, analyze or discuss the video (URL or local file).
- First published：2026-06-30
- data as of：
- 2026-09-30
- Licence：
- MIT
When to use it
When an agent needs to understand a video's content — tutorials, meeting recordings, product demos.
Caution
Depends on local tooling (ffmpeg-style) and a transcription model; long videos mean many keyframes and heavy context cost.
Install prompt
請幫我安裝「Claude Real Video」技能：從 https://github.com/HUANGCHIHHUNGLeo/claude-real-video/tree/master/skills/claude-real-video（技能路徑：skills/claude-real-video/） 取得完整的技能目錄，放進我的 Agent 的技能目錄（例如專案內的 skills/ 資料夾，或該 Agent 文件指明的全域技能位置），確認裡面的 SKILL.md 存在、frontmatter 的 name 與 description 完整，然後列出這個技能宣告的腳本、外部工具與 API 需求，先不要執行任何腳本。Paste this to your agent. It deliberately contains no tool subcommands — check the tool's official docs for commands.
Compatible with
Related patterns