check audio encoding and decoding

[markvdb]

We're currently using aac throughout the chain in voctomix.

According to Luc, opus might also be a good intermediary option for going to voctomix:

• low bandwidth
• low cpu requirements for encoding
• optimised for speech applications

To be checked:

• cpu usage on A20 Olinuxino Lime2.
• optimal codec & parameters?

Start with an uncompressed 44.1 KHz 16bit stereo input wav file.

Throw an audio stream at the Lime2 using ffmpeg on a different machine. Encode it live with ffmpeg using the -re option to get an idea of the realtime encoding performance/cpu requirements?

[libv]

Opus definitely would be great for low latency transmission to the other box in the room, so that the other box can play the audio over the PA system. Example project: http://www.pogo.org.uk/~mark/trx.git

Then, when opus is done with a buffer of, iirc, 512 frames, we can hand that buffer and the timestamp over to the AAC encoder for muxing in with the h.264 stream.