Video research

YouTube Transcript API

Turn a public YouTube video into timestamped text you can check against the source. Pull a passage for an SEO brief, competitor review, or research workflow and keep a link to the moment it came from.

Inspect the recorded result Read the API docs

Send a video ID

Choose the language and send the ID from your application.

curl https://mcpscraper.dev/youtube/transcribe \
  -H 'Content-Type: application/json' \
  -H 'x-api-key: YOUR_API_KEY' \
  --data '{"videoId":"wvJIOQndzow","language":"en"}'

In an MCP client, use youtube_transcribe with the same video ID.

Inspect the recorded result

We sent the video ID for Google Search Central’s “Introducing SEO Made Easy!” and received 21 timestamped chunks. REST returns plain text, Markdown, HTML, and a timestamp pair in seconds for each chunk. MCP exposes those bounds as startSec and endSec.

One selected chunk from the recorded REST response. Other chunks, fields and transcript text are omitted.

{
  "videoId": "wvJIOQndzow",
  "durationMs": 63130,
  "chunks": [
    {
      "timestamp": [
        20.699,
        25.019
      ]
    }
  ]
}

Three REST timestamp pairs compared with a separate machine transcription of source audio
Range (seconds)	Source	Audio check
`[5.24, 9.9]`	Play from 0:05	Inconclusive. The audio sample ended early.
`[20.699, 25.019]`	Play from 0:20	All seven words matched in order.
`[40.02, 44.34]`	Play from 0:40	Six of seven words matched consecutively.

We compared words after normalizing case and punctuation. Adjacent chunks can overlap. In this run, the last chunk ended at 63.130 seconds and the player reported 61.881 seconds. Keep the original range with your evidence and check the passage in the player before citing it.

Turn one passage into a research note

If you’re preparing beginner SEO training for a client, open the source at 0:20. The speaker describes the series as quick tips about search. Record that as your reason to include it in the brief, along with the video ID and the 20.699–25.019 second range.

If you’ll need it again, save and reopen the checked research in Memory Vault with the source link.

Check whether the API fits your workflow

Try a representative public video first. Transcript availability depends on the source and the retrieval path available for that video. If you need to download captions for a video you have permission to edit, use YouTube’s caption download API.

Record the video URL, required language, and whether you need existing captions or can use a transcript generated from the audio.
Choose one passage you would cite. Keep the start and end times in seconds and open the video at the start. Check the wording and language against playback.
Ask a teammate to repeat the playback check. If the language, wording, or timing fails the check, record why and leave the passage out of the brief.

Budget by video length. The YouTube transcription rate card lists a 200-Credit base plus 2 Credits per minute, reconciled to the actual video length.

Try a video from your own research

Start with one public video and check whether its timestamped chunks fit your brief or application.

See Starter pricing