Backup audio without camera stream

[yoe]

Currently, all audio goes through the camera stream. We have a primary audio stream that uses separate microphones and that goes through the camera in one channel, and a second audio stream that just uses the camera microphone in the back of the room through another channel. This is sent to the voctomix machines as a stereo audio stream.

The two audio channels are then separated into two audio streams, and stored as that on disk. We then allow reviewers to select the first or the the second stream in the review system, and also offer the possibility of selecting a mix of both channels (as a sort of "best of both worlds" situation).

This works fine for situations where the problem was with the microphones or the mixer, or something along those lines. It however means that the camera stream is a single point of failure, and if for some reason the camera stream drops during a talk (and this happens a few times during every FOSDEM), then unfortunately there is nothing that can be done.

We should evaluate the possibility of adding an ambient microphone to the front-of-room videobox (the one that captures slides), which can function as a last-resort audio stream. It should be muxed as a 3rd audio stream into the final recorded video. In cases of lost camera streams, this would still allow us to recover something. The quality will probably be terrible, but it would be better than nothing.