Quality wise whats the very best lossy codec?
Depends on the application... but maybe http://dev.gentoo.org/~redhatter/misc/mab135-project.pdf will help your decision process.
I actually did a comparison between MP3, Vorbis and AAC as a university project. Using analytical means, it was determined that for an AAC file at 128Kbps, Vorbis was roughly equivalent at 160Kbps, and MP3 was equivalent at 224Kbps.
I personally prefer AAC, but it's also because i actually LIKE using iTunes.
oog's pretty good, but it doesn't play everywhere.
WMA, WMApro, and WMA variable bit rate, I must say are pretty damn good as well. Just about anything that's come out since MP3 will be better.
Really?
I tried AAC at 128 kbps and it was horrible, 320 is decent though.
I've heard talk that Musepack (MPC) is the best at higher bitrates.
I swear that OGG Vorbis at it's highest setting sounds almost exactly like a lossless file.
>>7
I'd say that you're experiencing from the placebo effect. Do a double blind test of a mp3 @ 128kbps bitrate and the same mp3 @ 256kbps bitrate and you're unlikely to tell the difference.
>>8
i've done that, and i can tell the difference between 224 and 256 mp3. anything higher than 256 i can't tell any difference. but i use flac for everything except putting music on my phone because my phone can only play aac or mp3.
>>8 Actually, MP3 compression tends to fail hopelessly at higher frequencies...
Listen for distortion in percussion instruments like high hats/symbols, and in audience applause. Both of these contain many spectral components that are difficult to compress, giving rise to distortion.
I switched from VBR MP3 (160~192Kbps) to VBR Vorbis (112~128Kbps) and noticed an immediate difference in the sound quality. My later Fourier analysis of the codecs later confirmed what I was hearing.
>>2 Can you publish those full results?
I'm very surprised that AAC did better than Vorbis.
Were you using the latest aoTuV?
I wish WMP would be vorbis compatible.
I guess that's Micro$oft for you.
I wish iTunes would be vorbis compatible.
I guess that's Apple for you.
WMP is Vorbis compatible, by installing the codec. Microsoft is actually better than Apple in that regard. Fuckin' Apple.
High-Bitrate AAC for audio and h264 for video in an MKV container, with ASS subtitles if required, are the stuff dreams are made of.
>>15
vorbis is still superior as it is free.
Dumpster trash is superior to fine restaurant cuisine as it is free.
>>17
That is a shitty comparison. Vorbis and AAC are of comparable quality at high bitrates. In contrast, dumpster trash and fine restaurant cuisine stand far apart in quality, so the fact that one is free is not usually the deciding factor between the two.
> My later Fourier analysis of the codecs later confirmed what I was hearing.
The use of anything other than double-blind ABX tests have been pretty thoroughly debunked. Spectrograms are popular yet quite useless; your eyes and ears are different things.
Double-blind, yes, but ABX isn't necessarily the best way.
I can't find the paper now, but there's a test where you're given an interface to listen to three audio recordings: one is the reference and is labeled as such, and of the other two, one is the reference and the other has been mauled by lossy codecs in some way (randomly, without a label which is which). You give each of the two samples a score from 1 to 5 where 5 = indistinguishable from the reference, 4 = perceptible difference, 3 = slightly annoying artifacts, 2 = annoying artifacts, 1 = unlistenable garbage. (Thus if the sample that is exactly the same as the reference gets a 3, the results are suspect.) I think this is possibly better than ABX testing, and would like to spread the word about it.
>>17 was never saying that Vorbis was shit, they were saying that something being free isn't the only factor in deciding how good something is.
>>23
AAC and Vorbis are similar in technical capability. The thing that sets Vorbis apart from AAC is that it is free. That is why it is better.
Also, outside of Apple products (which admittedly are pretty much a monopoly, sadly, but still), Vorbis enjoys better hardware and software support.
>22
Nice idea, I think I'll try that.
It's called ABC-HR, I found out. A is the reference signal, and either B or C is the Hidden Reference.
horribly late, but h.264 would be my pick for overall size/quality
For video, definitely. The bulk of the discussion is talking about audio but the OP never specified that.
Haha, I didn't even think about video.
I guess video's more complex because you have to factor in how many dots are used (e.g. 720x480), while for audio, we all pretty much agree on 44100 dots per second (and for some reason, lossy codecs aren't used for the few high-resolution things like 24bit/96khz). Are there standard recommended settings for video encoders, same as audio encoders, that'll generally give "transparency"?
JPEG
For video I use h264 (well, the free x264 alternative, actually).
For audio... ogg or AAC (or AC3 if it is available/required).
In a MKV container, most probably.
Basically, what >>15 said.
Clarification, h264 is a codec, x264 is an implementation of the h264 codec. It isn't an alternative to it.
>>30
you can set alpha channels, but no transparency.
also, just for the sake of mentioning - 720x480 is a wrong aspect ratio for computer intended media. it's computer equivilent would be 640x480
720x480 would be correct for tv screens which have taller pixels (4:3 rather than a monitor's 1:1 PAR)
REAL MEDIA VARIABLE BITRATE :-p