Question 1

Does it transcribe the audio?

Accepted Answer

No. It reads the caption track YouTube already publishes, and there is no speech-to-text engine behind it. A video with no captions at all returns an error saying so, rather than a transcript that was guessed. That is the whole reason it costs one credit: a caption track is a few kilobytes of text, and running speech recognition over an hour of audio is not.

Question 2

How do I know if the transcript is any good?

Accepted Answer

The response says which track it came from. Captions written by whoever uploaded the video are punctuated and spelled correctly; YouTube's automatic captions are speech recognition, and they run words together, miss proper nouns and have no sentence breaks. Both are returned when they are what exists, and the X-Transcript-Source header and the source field say which you got.

Question 3

Is the video downloaded?

Accepted Answer

No, on any path. The request asks YouTube for the video's metadata and its caption track and stops there, which is why it comes back in a second or two rather than minutes. Nothing is written to disk and nothing is kept after the response.

Question 4

Can I get another language?

Accepted Answer

Any language the video has a caption track in. Ask for one it does not have and the answer is an error listing the ones it does, never a silent fall back to English — a transcript in the wrong language that does not say so is worse than no transcript.

Question 5

What does it cost?

Accepted Answer

One credit per video, flat, whatever its length. Credits are bought in packs, never expire, and buy any other operation here as well. There is no subscription.

YouTube transcript

Related