[Ffmpeg-devel] [PATCH] faad decoding error return

[Oded Shimon]

On Sun, Mar 05, 2006 at 02:56:22PM +0200, Oded Shimon wrote:
> On Sun, Mar 05, 2006 at 01:48:23PM +0100, Diego Biurrun wrote:
> > On Sun, Mar 05, 2006 at 12:33:55PM +0000, M?ns Rullg?rd wrote:
> > > 
> > > Any volunteers to write a proper AAC codec?
> > 
> > Now that would be wonderful..
> > 
> > Maybe we can try out an idea I've been having for a long time and put
> > out a bounty for an AAC implementation.  I'm willing to put, say, 50EUR
> > where my mouth is, if a few others join in maybe we can motivate
> > somebody to finally implement this.
> 
> Hmm, what is so horrible about FAAD? (I'm asking, I know nothing of this)
> 
> > Also see Mike's writeup about AAC in the wiki:
> > 
> > http://wiki.multimedia.cx/index.php?title=Understanding_AAC
> 
> Hmm, I can proudly say I now have at least some experience in audio 
> decoding, and I'm pretty bored now that NUT is either stalled or complete, 
> so maybe I can start working on this... Are there any (readable) specs 
> somewhere?

Just thought I'd give a status report (and brag) - I can (almost) 
successfully decode a AAC LC file created with faac! Decoder is very much a 
mess right now, and I'm not gonna bother posting it here cause it will only 
work with a very strict set of files (and even those not very well), but I 
can proudly say I've given my decoder a bitstream and got back human 
listenable audio... To further brag, it is almost 2 times faster than 
faad2, even though I made no attempt at optimization. But this might change 
when the deocder is actually complete...

A question to any knowledgable ones here - the spec says I can use 
global_gain with offset of -100 to get 16bit pcm result, but the 
global_gain in this file is ~170, which makes the values ~200,000, 
obviously unreasonable for 16bit pcm. With offset -140 files seems to be 
same volume as faad2. On another file I tested, I got jumps of global_gain 
between 100 and 150. Anyone know whats going on? I'll try figuring out from 
faad2 source...

- ods15