TranscriptYTHow to transcribe a YouTube video with Whisper
- 1
Copy the video link
Any public YouTube video, Short, or finished livestream.
- 2
Paste it above
If the video has captions you get them right away. If not, click Transcribe the audio with AI to run Whisper on the audio.
- 3
Download the result
Timestamped text, SRT, VTT, JSON-style segments through the API, or a PDF.
Whisper without the pipeline
The usual pipeline is yt-dlp to fetch the audio, ffmpeg to convert it, and Whisper on a GPU. This does all three from a link, with the audio fetch handled on our side so YouTube's download blocks aren't your problem.
Captions when they exist, Whisper when they don't
Running Whisper on a video that already has creator-uploaded captions wastes time and money. The tool checks for captions first and only transcribes the audio when there are none.
Need more?
Need transcripts at scale? Use the API
The same transcripts are available through the YouTube Transcript API: timestamped JSON, text, SRT, VTT, CSV, or Markdown, 150+ languages, and AI transcription for videos without captions. Or connect the YouTube MCP server to Claude, Cursor, or Codex and just paste links in chat.
curl "https://transcript-yt.com/v1/transcript?url=5f3Pn-N6yLc&format=text" \ -H "Authorization: Bearer ts_live_..."
Common questions
Which Whisper model is used?
Whisper large-v3 turbo, the fast version of OpenAI's most accurate open model, with the spoken language detected automatically.
Why not run Whisper myself?
You can: download the audio with yt-dlp and run Whisper on a GPU. In practice YouTube often blocks yt-dlp downloads from servers and laptops alike, and a two-hour video takes a long time on a CPU. Here it's one link and about two minutes.
How much does it cost?
Videos with captions are free. Whisper transcription is 1 credit per minute of audio on paid plans; the $3/month plan includes 1,000 credits.
Is Whisper better than YouTube's auto-captions?
Usually, for punctuation and capitalization, and for videos YouTube didn't caption at all. When a video has creator-uploaded captions, those are the most accurate and are used first.
Can I use it from code?
Yes. The API's /v1/transcript endpoint returns captions when they exist and falls back to Whisper when they don't, as JSON, text, SRT, or VTT.