From 85c8a405e3651dc041b758f4744d4fb3c4c55e01 Mon Sep 17 00:00:00 2001
From: bashonly <88596187+bashonly@users.noreply.github.com>
Date: Sun, 1 Jun 2025 18:09:47 -0500
Subject: [PATCH] [ie] Improve JSON LD thumbnails extraction (#13368)

Authored by: bashonly, doe1080

Co-authored-by: doe1080 <98906116+doe1080@users.noreply.github.com>
---
 test/test_InfoExtractor.py | 14 ++++++++++++++
 yt_dlp/extractor/common.py |  6 +++---
 2 files changed, 17 insertions(+), 3 deletions(-)

diff --git a/test/test_InfoExtractor.py b/test/test_InfoExtractor.py
index c6ff6209a8..bc89b2955e 100644
--- a/test/test_InfoExtractor.py
+++ b/test/test_InfoExtractor.py
@@ -314,6 +314,20 @@ class TestInfoExtractor(unittest.TestCase):
                 },
                 {},
             ),
+            (
+                # test thumbnail_url key without URL scheme
+                r'''
+<script type="application/ld+json">
+{
+"@context": "https://schema.org",
+"@type": "VideoObject",
+"thumbnail_url": "//www.nobelprize.org/images/12693-landscape-medium-gallery.jpg"
+}</script>''',
+                {
+                    'thumbnails': [{'url': 'https://www.nobelprize.org/images/12693-landscape-medium-gallery.jpg'}],
+                },
+                {},
+            ),
         ]
         for html, expected_dict, search_json_ld_kwargs in _TESTS:
             expect_dict(
diff --git a/yt_dlp/extractor/common.py b/yt_dlp/extractor/common.py
index d5607296df..1174bd4f5e 100644
--- a/yt_dlp/extractor/common.py
+++ b/yt_dlp/extractor/common.py
@@ -1675,9 +1675,9 @@ class InfoExtractor:
                 'ext': mimetype2ext(e.get('encodingFormat')),
                 'title': unescapeHTML(e.get('name')),
                 'description': unescapeHTML(e.get('description')),
-                'thumbnails': [{'url': unescapeHTML(url)}
-                               for url in variadic(traverse_obj(e, 'thumbnailUrl', 'thumbnailURL'))
-                               if url_or_none(url)],
+                'thumbnails': traverse_obj(e, (('thumbnailUrl', 'thumbnailURL', 'thumbnail_url'), (None, ...), {
+                    'url': ({str}, {unescapeHTML}, {self._proto_relative_url}, {url_or_none}),
+                })),
                 'duration': parse_duration(e.get('duration')),
                 'timestamp': unified_timestamp(e.get('uploadDate')),
                 # author can be an instance of 'Organization' or 'Person' types.