Docs

YoutubeToText developer documentation

The same engine the site runs on, as 4 REST endpoints and an MCP server. Queue a video, poll the job, read the text.

Three steps

  1. 1. Get a token. It lives on your account page and carries your minutes. Your API token.
  2. 2. Queue a video. The call answers straight away with a job id. The work runs after it.
  3. 3. Poll the job until its state is done or failed, then read the text off it.
curl
curl -X POST https://api.youtubetotext.ai/v1/api/transcribe \
  -H "Authorization: Bearer ytt_<your-token>" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ"}'

The pages

QuickstartToken, one request, one poll, your first transcript.AuthenticationBearer tokens, where they come from, and what a 401 means.API referenceEvery endpoint, its parameters and what comes back.Jobs and pollingThe twelve states, how often to poll, and reading the result.ErrorsStatus codes, the validation shape, and jobs that fail after 200.Minutes and limitsWhat a call costs in minutes, and every ceiling worth knowing.PythonA working script, and how this differs from reading captions.n8n, Make, ZapierThe same two calls, as nodes, with the wait in between.

What stays the same

Everything lives under /v1/ on https://api.youtubetotext.ai. Fields get added, and a job can pick up a new state, so read what you need and ignore the rest: today there are 12. Anything that would break a caller goes behind a new prefix, and this page says so before it happens.

The minutes are the same minutes as the site. A job queued here appears in your library, and one started in the browser can be read back through the API.