@@ -15,6 +15,25 @@ and a grainy film print never wanted the same one.
1515
1616![ videer processing a queue] ( thumb.png )
1717
18+ ## Changes in 3.13.3
19+
20+ - ** A subtitle track no longer takes the whole file down with it.** Encoding a subtitled MP4 to MKV failed
21+ before a single frame was written — the source's subtitles are ` mov_text ` , MKV was told to copy them, and
22+ Matroska is the one container that cannot hold ` mov_text ` . FFmpeg refused to write the header and the
23+ encode was over: ` Subtitle codec 94213 is not supported ` . Every file in a queue of MP4s went the same way.
24+ Each subtitle stream is now decided on its own — copied where the container accepts it, converted where it
25+ wants another text format (` mov_text ` to SRT for MKV, SRT/ASS to ` mov_text ` for MP4 and MOV), and left out
26+ only when it is a picture no conversion can place, which the log and the info panel both say out loud.
27+ - ** And if that judgement is wrong, the file still survives.** Knowing what a container will take means
28+ examining the source, and the examination can come back empty — no ffprobe on the machine, an input FFmpeg
29+ describes differently, a codec nobody anticipated. So FFmpeg's refusal is now treated as the answer the
30+ examination could not give: the encode runs again with the subtitles converted to the container's own text
31+ format, and if that is refused too, once more without them. A rejection happens while writing the header,
32+ so each retry costs about a second, and a file rescued by one is no longer reported as a file that failed.
33+ - ** The examination itself no longer needs ffprobe** : without it, the subtitle streams are read from the
34+ stream listing ` ffmpeg -i ` prints for any input it opens.
35+
36+
1837## Changes in 3.13.2
1938
2039- ** The Match Source Quality window flooded the console with errors.** It kept the chosen metric in an
0 commit comments