Transcript Kit

Turn a recording you own into a timestamped transcript with SRT and VTT captions, search what you have transcribed, find the best short clips in a talk, and translate the captions with a dub script.

Ask "transcribe this podcast episode", "make subtitles for my video", "where did I mention pricing?", "find the best 60 seconds of this talk" or "translate these subtitles to Spanish".

Transcript Kit works on audio and video the user owns or has the rights to use: a file uploaded to the chat, or a direct link to an audio or video file. It does not download from YouTube, TikTok, Douyin or any other platform, and it does not read pages of those sites. Each file can be up to 15 MB and 15 minutes long (mp3, m4a, mp4, wav, ogg, flac or webm).

Speech is transcribed with the Whisper large-v3-turbo model on Cloudflare Workers AI, which costs about $0.0005 per audio minute. It is free to use, with no sign-up. To keep it free, each user and the service as a whole can transcribe only a limited number of minutes per day (UTC); every answer says how many minutes are left.

Each transcript is kept for 24 hours so it can be searched and clipped, then deleted, and a library holds the latest 10. The media file itself is never kept. The search finds passages and repeated themes in your own transcripts only. Clip suggestions are text with start and end times and a hook line; no video is cut.

Captions can be translated into about 100 languages with the m2m100 model on Cloudflare Workers AI (about $0.0002 per 1,000 characters), keeping the original times, with a dub script of natural spoken lines sized to each segment. A call takes up to 18,000 characters, and translated text is not stored.

A transcript is a model's best reading of the audio and can contain mistakes, especially with names, accents or noise. Check it before you publish it.