Getting the Nepali video in
Drop in the video file unchanged: MP4, MOV, MKV, WEBM, AVI, M4V, 3GP or a phone screen recording. The sound track is pulled out on the server, so there is no convert-to-MP3 step, and every timestamp still matches the picture. If the video is online, paste its public link under Import from link: YouTube, Shorts, TikTok, public Vimeo, Dropbox or Google Drive share links, or a direct link to a video file. Only the audio is downloaded. Playlists and channels are refused, so queue videos one link at a time, and sign-in-only posts such as Instagram need to be downloaded and uploaded.
The draft opens beside a player that highlights the line being spoken. Click a line to replay it and correct the text right there.
Devanagari, Hindi and English mixing
Nepali is written in Devanagari, the same script as Hindi. The model has much more Hindi behind it, so auto-detect can mistake a Nepali video for Hindi, and the draft can slip into Hindi words. Keep Nepali selected, as it is on this page.
Expect a draft that needs careful correction rather than a finished text: verb endings, honorific forms (तँ, तिमी, तपाईं) and names are the usual fixes, and fast or noisy speech gives a rougher result. Kathmandu speech mixes in plenty of English, and Nepali speakers in Darjeeling, Sikkim and Bhutan, or working in the Gulf, Malaysia, South Korea and Japan, bring in words from where they live. Videos in Maithili, Bhojpuri, Newari or another language of Nepal will not transcribe well with Nepali selected.
Nepali videos people transcribe
Common uploads:
- YouTube podcasts and long interviews, for quotes and captions.
- News clips and political speeches.
- Loksewa and school exam-prep lectures, for notes.
- Family videos and messages sent home by relatives abroad.
- Your own TikTok and YouTube videos, for Nepali subtitles.
Nepali subtitles
Export SRT or VTT after editing. YouTube shows Devanagari correctly from the UTF-8 file; in desktop editors, pick a Devanagari font such as Noto Sans Devanagari or Mukta so conjunct letters join properly. DOCX and PDF suit a written copy, and CSV lists one line per row.
Tips and English
Use a close microphone, avoid music under speech, and switch on audio repair for street or crowd noise. Speaker labels for podcasts with two hosts are available on paid plans and with a Minute Pack.
The transcript stays in Nepali. For an English reading, open the Google Translate link on the transcript, or use the AI prompt export to hand the text to ChatGPT, Claude or Gemini. For audio-only recordings, see the Nepali audio to text page.