I think sending a PR with what you've got so far makes sense.

Heads up: We have some work planned that will change how subtitle data flows from muxed formats like MKV to the TextRenderer - and that will likely affect this issue too.

Some background: Currently we put samples in the SampleQueue containing data in the 'original' subtitle format, and then parse/decode it inside the TextRenderer. We recently added support for parsing the subtitle data before we put samples into the SampleQueue, which brings various advantages and simplifications. This currently only works for sideloaded subtitle files (those specified in MediaItem.localConfiguration.subtitleConfigurations), and is opt-in (controlled by DefaultMediaSourceFactory#experimentalUseProgressiveMediaSourceForSubtitles). We are planning to standardise all subtitle-processing to follow this model, which (among other changes) will involve changing MatroskaExtractor to parse the subtitle data it reads from each sample as it goes along.

If you get the PR in before we start that work in MatroskaExtractor then it will help to ensure we don't somehow make this improvement completely impossible in the new architecture :)

I'm going to mark this as low priority to indicate that we (exoplayer-team) have no immediate plans to directly work on implementing this - but we will of course prioritise reviewing and merging any PRs you send us!

I would also like this feature. yt-dlp, for example, downloads all subtitles in vtt format. I have the following options for watching the videos with subtitles:

1. convert subtitles to .srt with ffmpeg.
2. use VLC Media Player.

Both options are not really nice.