it might be worth to take a look at Faac* (ups, see below) again and compare it to FDK AAC

http://ffmpeg.org/index.html#aac_encoder_stable
They flagged it as stable yesterday
https://github.com/FFmpeg/FFmpeg/commit ... 87f2777151
Apparently two developers pushed the development a lot this year (since GSoC)
Just found out about it and thought it might be worth sharing.
EDIT:
* just noticed that this isn't about faac but FFmpeg's native aac encoder
so sorry if I caused any confusion
Anyways, I just found out they are planning to remove faac in the future and offer native aac instead
https://trac.ffmpeg.org/ticket/2686#comment:432
faac will probably be scrapped, fast will have to be rewritten, and anmr is a big question mark at present. I've been working on ANMR and some problems have surfaced that don't seem easy to resolve, or at all possible with ANMR's approach. Surely it can be made not to crash, but beyond that I'm unsure how far we can push ANMR.
For now, the priority is twoloop.