X Video Transcribe
Transcribe X/Twitter video posts to text. Downloads the video, extracts audio, and uses Gemini for accurate transcription.
When to use
- User shares an X/Twitter URL containing a video
- User asks to "transcribe this tweet" or "what does this video say"
- User wants a summary of an X video
Usage
# Basic transcript
~/agent-workspace/skills/x-video-transcribe/scripts/transcribe.sh "https://x.com/user/status/123"
# With summary
~/agent-workspace/skills/x-video-transcribe/scripts/transcribe.sh "https://x.com/user/status/123" --summary
# Save to file
~/agent-workspace/skills/x-video-transcribe/scripts/transcribe.sh "https://x.com/user/status/123" --summary --output /tmp/transcript.md
Pipeline
- bird CLI — Fetches tweet JSON, extracts video URL
- curl — Downloads the video MP4
- ffmpeg — Extracts audio as MP3 (much smaller than video)
- Gemini API — Uploads audio, transcribes with gemini-2.0-flash
Requirements
birdCLI with auth (cookies in~/agent-workspace/secrets/bird.env)ffmpegGEMINI_API_KEYenvironment variable
Environment
| Variable | Default | Description |
|---|---|---|
GEMINI_API_KEY |
(required) | Google Gemini API key |
BIRD_ENV |
~/agent-workspace/secrets/bird.env |
Path to bird credentials |
GEMINI_MODEL |
gemini-2.0-flash |
Gemini model for transcription |