How it works
Drop in the video file without converting it: MP4, MOV, MKV, WEBM, AVI, 3GP or a screen recording from your phone. The audio is pulled out on our server, so every timestamp still points to the right moment in the video. If the video is online, choose Import from link and paste a public link from YouTube, TikTok, Vimeo, Dropbox or Google Drive, or a direct link to the file; only the sound is downloaded.
The draft opens beside a player. Click a line to replay it as often as you need; corrections are saved as you type.
A very rough starting point
Tibetan has extremely little training data in the speech model, so be prepared for a rough draft that needs heavy correction: wrong syllables, missing tsheg marks between syllables and words the model simply guesses. It is still useful for its timestamps and line breaks, which give you a timed frame to type the right text into. Slow, clear speech by one person near the microphone works best.
Spoken Lhasa Tibetan and the Amdo and Kham dialects differ a lot, and the model handles none of them well. Classical Tibetan recitation, chanting and prayers with ritual instruments are the hardest of all.
Teachings with interpreters
Many Tibetan videos are teachings where a lama speaks in Tibetan and an interpreter repeats each passage in English or Chinese. If the interpretation is what you need, select English or Chinese instead: those parts will come out far better, and you only need to mark or fix the Tibetan passages. If you need the Tibetan itself, split the recording if you can, so the Tibetan sections are transcribed on their own.
- Dharma teachings and public talks recorded in Dharamsala or Nepal.
- Interviews and oral histories with elders in exile communities.
- News clips and community meetings.
- Language lessons for Tibetan learners.
Tibetan script in subtitles
SRT and VTT exports are UTF-8 Tibetan. Stacked letters need a font with full Tibetan support, such as Noto Serif Tibetan or Jomolhari; without one, many editors show boxes or broken stacks. YouTube shows Tibetan subtitles when the viewer's device has a Tibetan font. DOCX and PDF keep the script for printing.
English and the audio page
The transcript stays in Tibetan; there is no built-in translation. To read it in English, open the Google Translate link on the transcript, or use the AI prompt export to hand the text to ChatGPT, Claude or Gemini. Machine translation of Tibetan is unreliable, so use it only as a hint. For audio-only recordings, see the Tibetan audio to text page.