How to filter VTT text tracks in AndroidX Media 1.4.0?

[jpgpuyo]

Hello,

I have detected a different behaviour between AndroidX Media 1.3.0 and AndroidX Media 1.4.0.
I would like to filter VTT text tracks.

Below I show a sample code to explain this question.

In AndroidX media 1.3.0, I'm doing it in the following way:

override fun onTracksChanged(tracks: Tracks) {
...
tracks.groups.filter{ 
    it.type == C.TRACK_TYPE_TEXT && 
    it.mediaTrackGroup.getFormat(0).sampleMimeType == "text/vtt" }
...
}

Now, in version AndroidX Media 1.4.0, I have to do the following:

override fun onTracksChanged(tracks: Tracks) {
...
tracks.groups.filter{ 
    it.type == C.TRACK_TYPE_TEXT && 
    it.mediaTrackGroup.getFormat(0).sampleMimeType == "application/x-media3-cues"  &&
     it.mediaTrackGroup.getFormat(0).codecs == "text/vtt"
}
...
}

In this particular case, we are not using the Exoplayer subtitles view, we are using our custom view, so I need to retrieve VTT tracks. I have checked release notes but I am not sure where this change is explained.

Is it the expected behaviour or this should be reported as a new bug?
Which is the right way to filter VTT text tracks?

Thanks!

[icbaker]

Is it the expected behaviour or this should be reported as a new bug?

The change in MIME type is expected - because from 1.4.0 onwards the bytes written into the SampleQueue for a WebVTT subtitle track are no longer WebVTT UTF-8 text, but instead a media3-specific serialization of the subtitle data - hence the change in MIME type (because the parsing logic for these bytes also needs to change).

---

In this particular case, we are not using the Exoplayer subtitles view, we are using our custom view, so I need to retrieve VTT tracks. I have checked release notes but I am not sure where this change is explained.

Could you elaborate on this part a bit more - are you parsing the WebVTT data yourself (so you want to consume the UTF-8 text data directly), or are you consuming Cue objects from Player.Listener.onCues or similar? If you are consuming the Cue objects, could you add some more detail on why you care about the 'original' subtitle format (since the Cue object is format-agnostic)? This will help me understand the best way to help you.