fix: YouTubeConverter.accepts() silently drops youtu.be short URLs

[kuishou68]

Bug Description

The YouTubeConverter.accepts() method in _youtube_converter.py only matches URLs that begin with https://www.youtube.com/watch? and silently rejects all youtu.be short URLs (e.g. https://youtu.be/dQw4w9WgXcQ). This means short YouTube links are never converted — the converter returns False and falls through to a generic HTML converter, producing noisy output instead of a clean transcript.

Reproduction

from markitdown import MarkItDown
md = MarkItDown()
# This works:
result = md.convert("https://www.youtube.com/watch?v=dQw4w9WgXcQ")
# This does NOT go through YouTubeConverter:
result = md.convert("https://youtu.be/dQw4w9WgXcQ")

Root Cause

In packages/markitdown/src/markitdown/converters/_youtube_converter.py, the accepts() method contains:

if not url.startswith("https://www.youtube.com/watch?"):
    # Not a YouTube URL
    return False

This check must also accept https://youtu.be/<id> and https://www.youtube.com/shorts/<id> URLs.

Expected Fix

The accepts() check should be extended to recognise short URLs and extract the video ID from the youtu.be/<id> path segment, and the convert() method should handle extracting the video ID from both URL forms.

Related

Issue #1704 mentions that short URLs (youtu.be/…) are silently skipped.

[kuishou68]

I've opened PR #1731 to fix this issue. The fix introduces a shared helper function _extract_youtube_video_id(url) that correctly handles youtu.be/<id> short URLs, youtube.com/shorts/<id> URLs, and the original youtube.com/watch?v=<id> form. Both accepts() and convert() now use this helper, so short-URL transcripts are fetched correctly.

[Jah-yee]

Good day,

I've prepared a fix for this issue. The problem was that YouTubeConverter.accepts() only matched full youtube.com/watch URLs, not short URLs like youtu.be or youtube.com/shorts.

Solution:

1. Extended accepts() to recognize all YouTube URL patterns
2. Added _extract_video_id() helper to extract video ID from any URL format
3. Updated convert() to use the helper for transcript fetching

Complete patch: https://gist.github.com/Jah-yee/1c51d1f327e2d28244a8aa197f296160

Would you like me to submit this as a PR?

感谢你们的奉献，希望能提供帮助。如果我解决得有问题或有待商妥的地方，请在下面留言，我会来处理。

Warmly,
RoomWithOutRoof