Twitter/X: Unpredictable behaviour following links in Tweets

[leonbeon]

Twitter/X: Unpredictable behaviour following links in Tweets

[bashonly]

Twitter has two different concepts of embedded URLs: a "card", and a pure link. The Meta tweet is an example of a "card": it has a large image (the card) that is also clickable as a link to the content. Your other two examples are pure links.

yt-dlp's Twitter extractor prioritizes cards over pure links, since cards can also be actual embedded videos, Twitter broadcasts, Spaces, etc. It tries to extract the card from the Meta tweet, but it doesn't recognize the card "name", so it returns no formats found (i.e. extracts tweet metadata without a video or URL result).

Adding the missing card name to the extractor's code would fix the inconsistency:

diff --git a/yt_dlp/extractor/twitter.py b/yt_dlp/extractor/twitter.py
index ecc865655..cfd34eff4 100644
--- a/yt_dlp/extractor/twitter.py
+++ b/yt_dlp/extractor/twitter.py
@@ -1499,7 +1499,7 @@ def get_binding_value(k):
                     'url': f'https://twitter.com/i/spaces/{get_binding_value("id")}',
                     'ie_key': TwitterSpacesIE.ie_key(),
                 }
-            elif card_name == 'summary':
+            elif card_name in ('summary', 'summary_large_image'):
                 yield {
                     '_type': 'url',
                     'url': get_binding_value('card_url'),

However, some more investigation should probably be done on other potential card names to be included in this tuple.

I do get the sense that maybe what you really want is to be able to extract tweet metadata for all tweets instead of following any of these redirects. It should be stated that yt-dlp is first and foremost an audio/video downloader, not a metadata extractor. If no video is embedded in the tweet, yt-dlp will (and should) follow the embedded URL to try to find any video at its destination. The ability to extract pure tweet metadata was added to the Twitter extractor as a bonus; it is not the focus. But if you want to be able to extract only twitter metadata from every tweet, this is possible via the yt-dlp Python API and the extract_flat param:

import json
import yt_dlp

URL = 'https://twitter.com/YouTube/status/1780725427520065913'

ydl_opts = {
    'extract_flat': True,
    'ignore_no_formats_error': True,
    'no_warnings': True,
    'quiet': True,
}

with yt_dlp.YoutubeDL(ydl_opts) as ydl:
    info = ydl.extract_info(URL, download=False)
    info = ydl.sanitize_info(info)

print(json.dumps(info))