Closed itzexor closed 1 year ago
Hi and thanks for the detailed report 👍
I found the issue and wrote a more robust piece of code to handle Vorbis chapters. ATL was indeed expecting 1-indexed chapters only.
The fix will be available with next release.
Fix available in today's v4.18
Hello, I'm the reporter from the other issue you were talking to @sandreas about recently. During my work in figuring out what the general behavior should be with those kind of files, I found that ffmpeg automatically rearranges any vorbis chapter metadata that it finds so that it starts at index 0.
Once I tested transforming my file with ffmpeg in order to see what it decided to do with the "out of bounds" chapter data, it no longer worked with tone. I started getting this backtrace:
The output shown above is from a super basic file I crafted for the test using this command with opusenc:
$ opusenc --comment CHAPTER000=00:00:00.000 --comment CHAPTER000NAME="Chapter 1" --comment CHAPTER001=00:00:07.000 --comment CHAPTER001NAME="Chapter 2" no_sound_test.wav no_sound_test_000_chapter.opus
however changing that command to simply use chapter001 and chapter002 instead works fine:
That page I referenced in the other issue (https://wiki.xiph.org/Chapter_Extension) shows their example starting at index 1:
CHAPTER001=00:00:00.000
and this is the format I've been putting into my files the whole time so I hadn't noticed. After encountering this issue, and going back to that page, I see it does specify the range as index 000-999.I guess maybe atl is expecting that vorbistag chapters start at index 1 or something? To be extra clear this is occurring in as far as I can tell completely valid ogg opus files, I haven't done any weird chapter length trickery to these test files.
Here's the wav and the resulting file if you need them: no_sound_test.zip