following a discussion on IRC, I thought I would move this here so it is more async.

The need is to be able to use Nuance/IBM Watson/Google Speech APIs in streaming mode for real-time transcription (essentially, captioning a video conference and getting a transcript of the proceedings).

My first idea was to simply get the frames from a recording as it is being written, though there might be some obstacles there.

It was suggested the media bugs infrastructure would work, at the C level, though I need to check which formats it supports.

Do you have any other suggestions?

Thank you!

Luca