How it works
What this tool actually does
Facebook to Transcript has two jobs: read captions Facebook already attached to a public video, or listen to a slice of audio and type what was said. It does not log into your account, and it cannot open friends-only posts.
1. You paste a public URL
Accepted hosts: facebook.com, m.facebook.com, and fb.watch. Useful path shapes include /watch, /reel/, /reels/, /videos/, /share/v/, and /share/r/. If the share dialog says only friends can view it, stop — Extract and AI will both fail.
2. Extract reads captions
Many Watch videos and some Reels ship with automatic or creator-uploaded captions. Extract downloads those timed lines and turns them into TXT, SRT, and VTT. It is the cheaper, faster path. If Facebook never generated captions, Extract returns a clear error instead of inventing text.
3. AI transcribes audio
AI fetches the soundtrack (not a full 4K file), trims it to the 15-minute range you picked, and sends that clip to a speech-to-text model. Pick a language when you know it — auto-detect is a fallback, not a magic trick. Music-only clips and heavy crowd noise produce poor results. That is expected.
4. Auto is Extract, then AI
Auto tries captions first. If none exist, it falls through to AI (when an API key is configured on the server). Use Extract if you specifically want Facebook’s own text. Use AI if those captions are in the wrong language or full of junk.
What we refuse
- Private, login-gated, or expired share links
- Messenger / chat history (use Facebook’s data download)
- Copyrighted material you do not have a right to process
- Bulk scraping and bot farms — rate limits apply