Video, image & audio generation
bradautomates/claude-video
Gives Claude the ability to watch and transcribe any video
- 16k
- MIT
- Python
What it is and why it matters
claude-video is a skill that lets Claude ingest video (download it, extract frames and transcribe), a capability it lacks out of the box. It combines the transcript with frames selected by the chosen mode, from transcript-only to broad frame capture, to balance cost and detail. If the video has no transcript, it generates one with the Whisper model, avoiding routing through paid external services.
Key points
- Downloads the video, extracts frames and hands them to Claude alongside the transcript
- Four modes control how many frames are captured, from transcript-only to uncapped, by cost and detail
- If the transcript is missing it generates one with Whisper; useful to learn from videos and make them searchable
Where we saw it
Chase AI 2026-07-08 Presented as the tool that gives Claude the ability to watch video, which it lacks out of the box. It explains the four modes (transcript, efficient, balance, token burner) that adjust captured frames and the use of Whisper when there is no transcript. 5 Open Source Repos That Fix 95% of Claude Code's Problems The Next New Thing 2026-07-16 The hosts highlight that it lets the agent see and hear video, edit it and above all learn from it, making already-watched material searchable; they mention repurposing long recordings by crossing transcript and visual moments. Free: AI design tools, tokenmaxxing, & more top GitHub repos
Same category