Good words deserve a second look.
A useful sentence can disappear in the time it takes to reach for a notebook. WatchToText is a small tool site, free for most videos, for bringing those words back into view.
Paste a public video link from YouTube, TikTok, or Instagram. Read the transcript, copy a passage, or save a TXT or SRT file. TikTok and Instagram also support VTT.
FROM SCREEN TO PAGE
Two routes to the words.
01 / Read the existing captions
YouTube uses available caption tracks. TikTok also reads a caption track when one is available, whether uploaded by its creator or generated by the platform. The transcript reflects those captions, including any errors already in them.
02 / Listen to the audio
For TikTok without a caption track, and for Instagram videos, the site extracts the video’s audio and uses speech recognition to turn it into text. YouTube videos without captions do not use this fallback. This takes longer, and machine transcription can get words wrong.
The result tells you which source produced the text. If a video cannot be read, the status message explains which step failed.
Less between you and the text.
- No account, sign-up, or browser extension.
- No per-user link history on the server, and exported files are created in your browser. Public YouTube captions and TikTok/Instagram transcripts may be cached for up to 24 hours and reused for the same video. Some YouTube responses without transcript text may be cached for up to 6 hours. Uploaded audio, video and transcript text are not stored. Our Privacy Policy explains the cache and usage counters.
- No bulk uploads or job queues. Every export format is free. The one paid option is a one-time credit pack for YouTube videos that block free tools.
If no usable captions or recognisable speech can be found, the tool reports that instead of filling the page with invented text. Speech recognition can still make mistakes, so check important quotations against the video.
Behind the pages.
WatchToText runs on Cloudflare Pages. Server functions fetch captions or audio and, when needed, call speech recognition on Workers AI.
Pipeline notes, test results, and a small command-line transcript tool are available in our public notes repository ↗.
The site is run by the team behind the WatchToText name. We do not publish personal names, addresses, or phone numbers. For a question or correction, send us a note. For details about data and analytics, read our privacy policy.