HydrogenAudio

Lossy Audio Compression => Ogg Vorbis => Ogg Vorbis - General => Topic started by: dsimcha on 2010-03-18 14:57:15

Title: Can -q2 really be "almost" transparent?
Post by: dsimcha on 2010-03-18 14:57:15
I'm sitting here listening to some -q2 encoded Vorbis (using aoTuv Beta 5.7; nominal bitrate 96 kbps) and, after thinking for years and years that nothing below 128 kbps could possibly sound decent on any codec (based on listening to MP3, WMA and AAC) I amazingly find that I can't reliably ABX it from the original FLACs.  I can ABX some samples, but the artifacts are really, really minute things that I would never pick up in more casual listening and there are plenty of samples that I can't ABX at -q2 that I can ABX at substantially higher bitrates with other codecs.  Therefore, I call -q2 "almost" transparent.

Have others had this experience or does my hearing or equipment just suck?  Also, has anyone ever seen any other codec besides aoTuv that gets even remotely close to transparency in the ~96 kbps area?  I'm not aware of any, which is why I'm so shocked by these results.
Title: Can -q2 really be "almost" transparent?
Post by: timcupery on 2010-03-18 15:07:15
Depends a lot on the person, more than the equipment. Maybe your hearing sucks. But aoTuv 5.7 is pretty good, too.
To judge the "best" codec overall, you need listening tests with a large sample of listeners. The codec that performs the best overall probably will not sound the best to all listeners.
Different codecs produce somewhat different artifacts, and people aren't all sensitive to the same artifact.

So, since we're talking about you and your hearing here, I'd be interested how well you can ABX aac files at a similar bitrate. Or even mp3...
Title: Can -q2 really be "almost" transparent?
Post by: dsimcha on 2010-03-18 15:12:43
Depends a lot on the person, more than the equipment. Maybe your hearing sucks. But aoTuv 5.7 is pretty good, too.
To judge the "best" codec overall, you need listening tests with a large sample of listeners. The codec that performs the best overall probably will not sound the best to all listeners.
Different codecs produce somewhat different artifacts, and people aren't all sensitive to the same artifact.

So, since we're talking about you and your hearing here, I'd be interested how well you can ABX aac files at a similar bitrate. Or even mp3...


At similar bitrates, MP3 and AAC are just painful to listen to, so I've never even tried to ABX it.  At higher bitrates, I can easily ABX both LAME VBR and Nero AAC in the 128 kbps range.  (IIRC -V6 for LAME and -Q0.38 for AAC, though the average bitrates were definitely above 130 for both on the samples I used).  I've ABXed -V5 a few times with recent versions of LAME, but it isn't easy. 

I find that I tend to be extremely sensitive to the washed-out, ringing sound that MP3 and AAC produce in music with lots of high frequency content.  I can hear the artifacts of -V6 LAME even in very casual listening.  OTOH I find that I'm much less sensitive to distortions of the stereo image or to Gaussian/analog-like noise.
Title: Can -q2 really be "almost" transparent?
Post by: felix26591 on 2010-03-18 15:25:46
Depends a lot on the person, more than the equipment. Maybe your hearing sucks. But aoTuv 5.7 is pretty good, too.
To judge the "best" codec overall, you need listening tests with a large sample of listeners. The codec that performs the best overall probably will not sound the best to all listeners.
Different codecs produce somewhat different artifacts, and people aren't all sensitive to the same artifact.

So, since we're talking about you and your hearing here, I'd be interested how well you can ABX aac files at a similar bitrate. Or even mp3...


At similar bitrates, MP3 and AAC are just painful to listen to, so I've never even tried to ABX it.  At higher bitrates, I can easily ABX both LAME VBR and Nero AAC in the 128 kbps range.  (IIRC -V6 for LAME and -Q0.38 for AAC, though the average bitrates were definitely above 130 for both on the samples I used).  I've ABXed -V5 a few times with recent versions of LAME, but it isn't easy. 

I find that I tend to be extremely sensitive to the washed-out, ringing sound that MP3 and AAC produce in music with lots of high frequency content.  I can hear the artifacts of -V6 LAME even in very casual listening.  OTOH I find that I'm much less sensitive to distortions of the stereo image or to Gaussian/analog-like noise.



Actually I think the same, been using vorbis aotuv b5.7 for at least a year, and at low bitrates I find it to be the best lossy encoder out there. I can listen to good music no washout maybe a few artifacts here and there, at quality 2 or 3 and when I do the same with AAC or MP3 I hear the difference, but aoyumi has made a great diference to the vorbis codec.
Title: Can -q2 really be "almost" transparent?
Post by: X-Fi6 on 2010-06-07 11:06:00
It depends on what part of the sound the codec attacks and how well you'll be able to hear it through your ear buds when you're listening to it from your portable media player (you're not really going to play these back from your 1TB hard drive on your computer, are ya? ).

Aoyumi's Vorbis is tuned to preserve the power of loud notes and the echos/ringing of the sonic landscape by muting bass and dumbing down ultra high frequency instead (among casing a better RLE ).

This works perfectly on songs like Scar Tissue (I settled for -q0 on that) or John Mayer - No Such Thing (same thing), but if you try it on another fast-paced song with lots of quick-muting hi-hats like Billy Talent - Try Honesty, it's awful unless you use at least -q4.