[utils] Use bytes-like objects in dfxp2srt

This fixes handling of non-UTF8 TTML subtitles

Closes #14191

1 files changed, 11 insertions, 7 deletions

diff --git a/youtube_dl/utils.py b/youtube_dl/utils.py
index 9e4492d40..b724e0b70 100644
--- a/youtube_dl/utils.py
+++ b/youtube_dl/utils.py
@@ -2572,14 +2572,18 @@ def srt_subtitles_timecode(seconds):
 
 
 def dfxp2srt(dfxp_data):
+    '''
+    @param dfxp_data A bytes-like object containing DFXP data
+    @returns A unicode object containing converted SRT data
+    '''
     LEGACY_NAMESPACES = (
-        ('http://www.w3.org/ns/ttml', [
-            'http://www.w3.org/2004/11/ttaf1',
-            'http://www.w3.org/2006/04/ttaf1',
-            'http://www.w3.org/2006/10/ttaf1',
+        (b'http://www.w3.org/ns/ttml', [
+            b'http://www.w3.org/2004/11/ttaf1',
+            b'http://www.w3.org/2006/04/ttaf1',
+            b'http://www.w3.org/2006/10/ttaf1',
         ]),
-        ('http://www.w3.org/ns/ttml#styling', [
-            'http://www.w3.org/ns/ttml#style',
+        (b'http://www.w3.org/ns/ttml#styling', [
+            b'http://www.w3.org/ns/ttml#style',
         ]),
     )
 
@@ -2674,7 +2678,7 @@ def dfxp2srt(dfxp_data):
         for ns in v:
             dfxp_data = dfxp_data.replace(ns, k)
 
-    dfxp = compat_etree_fromstring(dfxp_data.encode('utf-8'))
+    dfxp = compat_etree_fromstring(dfxp_data)
     out = []
     paras = dfxp.findall(_x('.//ttml:p')) or dfxp.findall('.//p')