Get a YouTube transcript in Python

Create a token on your account page and put it in an environment variable. The script below queues a video, waits for it, and prints the text.

import time
import requests

BASE = "https://api.youtubetotext.ai"
HEADERS = {"Authorization": "Bearer ytt_<your-token>"}

# Queue the video. This answers straight away with an id; the work runs after.
job = requests.post(
    f"{BASE}/v1/api/transcribe",
    headers=HEADERS,
    json={"url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ"},
).json()

# Poll until the job stops. Every 3 seconds is plenty.
while True:
    result = requests.get(
        f"{BASE}/v1/api/transcription/{job['id']}", headers=HEADERS
    ).json()
    if result["state"] in ("done", "failed"):
        break
    time.sleep(3)

if result["state"] == "failed":
    raise SystemExit(result["error"])

print(result["txt"])

How this differs from reading captions

The popular Python packages fetch the caption track YouTube already holds. That is fast and free, and it stops at the same place every time: a video with captions turned off has nothing to read, and an auto-caption track is what the uploader was given, not what you would get from a fresh transcription.

This API transcribes the audio instead. It works on a video with no captions at all, it puts in punctuation and paragraphs, it labels the speakers, and it costs minutes, which a caption fetch does not. If the video you want already has a good caption track and you only need the words, a caption library is the cheaper tool and we would rather say so. Our own measured accuracy is on the accuracy page.