Phaedras@gmx.net
2002-Jun-20 03:36 UTC
[vorbis] When will quality increase be unnoticable?
I started thinking about this after doing a little testing with AAC, MP3 and Ogg Vorbis. I was comparing the different formats at similiar bitrates. After a while I finally realized that they all sound more or less the same to me.>From 160 kbps on, I usually cannot detect any difference between a lossyencoding and the original source. If LAME is used, I have to strain to notice anything at 128. Ogg Vorbis is the same at slightly lower bitrates. I know there are enough people out there who, even without ABXing, could tell me exactly what sounds different, but personally I have a hard time to discern any changes made by a lossy encoder (at a decent bitrate) to a piece of music. I also believe the majority of people feel this way. And so I wonder: When will codec development stop concentrating on quality and start concentrating on size? When I think about where Ogg Vorbis is right now and where it will be at 1.0, I don't understand why one would even need these "discrete wavelets" that are being discussed. Artifacts, like those that occur with MP3 seem nonexistant with Vorbis. I don't notice any high-frequency "squishyness" How much more of a quality gain can even be achieved? -- GMX - Die Kommunikationsplattform im Internet. http://www.gmx.net <p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
Phaedras@gmx.net wrote:> When will codec development stop concentrating on quality and start > concentrating on size? When I think about where Ogg Vorbis is right now and where it > will be at 1.0, I don't understand why one would even need these "discrete > wavelets" that are being discussed. Artifacts, like those that occur with MP3 > seem nonexistant with Vorbis. I don't notice any high-frequency "squishyness" > How much more of a quality gain can even be achieved?The answer to your first question is 'never,' but it should be pointed out that the question is ill-posed to begin with. Quality and File Size are two (somewhat) correlated variables, but as the encoder becomes better, the correlation between sound quality and file size changes. With newer encoders you can achieve the same quality at a smaller file size, or you can enjoy better quality at the same file size. So, to assert that Vorbis development concentrates on one versus the other is quite untrue, because the two are inextricably linked to one another. These "wavelets" are an additional technique for encoding sound, and they are "needed" to further enhance the Quality/Size "ratio" that the encoder achieves. Vorbis users have to untrain themselves from the MP3 thinking that a specific bitrate implies a specific sound quality. (Heck, it's not even true for MP3 if you compare the different MP3 encoders that are out there!) With Vorbis, you specify the quality that you want to obtain. If, for example, -q3 is satisfactory to you now, it will remain satisfactory to you throughout future releases of the encoder (assuming of course that your listening preferences don't change in the meantime). But while the current encoder averages at approximately 112kbps to achieve -q3, future encoders will cut the bitrate down, and the files will become smaller. I hope this helps, Carsten Haese (Vorbis fan & advocate) --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
Phaedras@gmx.net wrote:>I started thinking about this after doing a little testing with AAC, MP3 and >Ogg Vorbis. I was comparing the different formats at similiar bitrates. >After a while I finally realized that they all sound more or less the same to me. >>From 160 kbps on, I usually cannot detect any difference between a lossy >encoding and the original source. If LAME is used, I have to strain to notice >anything at 128. Ogg Vorbis is the same at slightly lower bitrates. > >I know there are enough people out there who, even without ABXing, could >tell me exactly what sounds different, but personally I have a hard time to >discern any changes made by a lossy encoder (at a decent bitrate) to a piece of >music. I also believe the majority of people feel this way. And so I wonder: >When will codec development stop concentrating on quality and start >concentrating on size? When I think about where Ogg Vorbis is right now and where it >will be at 1.0, I don't understand why one would even need these "discrete >wavelets" that are being discussed. Artifacts, like those that occur with MP3 >seem nonexistant with Vorbis. I don't notice any high-frequency "squishyness" >How much more of a quality gain can even be achieved? > > >Concerning 'Quality vs Size': 1. I don't really think there should be much more decrease in size, because, after all, no matter what kind of technology you use or how advanced audio coding algorithms you use, one needs to face the fact that lots of the original audio information is removed, and can never be restored 100%. Too much at 128kbps, if you ask me, and this is, for me, the absolute lowest acceptable bitrate (think of how much is really removed: 128 reduces the data amount to 1/12th of the original size), and audio quality is, no matter which codec (though ogg beats'em all) rather poor below 128 kbps. (This is easy to hear if you have decent sound equipment/a decent stereo with OK speakers and, of course, the sound card in the computer has a lot to do with the final quality.) 2. Harddisks (and storage in general) aren't exactly getting any smaller these days, and the same with internet/network bandwith. So why the need to reduce SO much, losing audio quality on the way. I think compression ratios will go the other way in the future, to preserve more of the original sound. I think that audio codec developers should concentrate on getting more quality out of the standard ratios we have today, rather than trying to get the bitrate even lower. I would also like to point out that I am talking about encoding music here, not speech (or perhaps streaming of music...) or anything else where quality is not the number one priority. To me, quality is very important concerning music, and therefore I encode all my music in OGG format with an oggenc quality level of 8 (~256kbps). Some might call me a quality freak, but this is still my subjective opinion, comments are very welcome. I am quite interested in digital audio, though I have no background developing such things, but I know the general principles. Regards, Øyvind Stegard <p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
Phaedras@gmx.net wrote:> > I know there are enough people out there who, even without ABXing, could > tell me exactly what sounds different, but personally I have a hard time to > discern any changes made by a lossy encoder (at a decent bitrate) to a piece of > music.You're lucky...try not to learn. <p>> When will codec development stop concentrating on quality and start> concentrating on size?It's the same thing. If you improve quality you can decrease the size. <p><p><p><p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
i can hear differences between the wav and the ogg at -q-4, but not at -q-5. i haven't tried -q-4.99, since that is the highest level before lossless channel coupling. Andy ¸_./'\._¸ ¸.·¤**¤·.¸.·¤**¤·.. *·. .·* *Andy Dale* /.·*·.\*.·¤**¤·.¸¸.·¤**¤·.* andycool22@peoplepc.com AndyCool22 on AIM ICQ # 76178574 MSN Messenger andycool22@msn.com Yahoo! Messenger - andycool22 ________________________________________________ PeoplePC: It's for people. And it's just smart. http://www.peoplepc.com --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
many people on this list have recommended finding one you are happy with, then jumping up one, like if you like 3, use 4, in case your ears get better you don't want to reencode and rerip everything. Andy ¸_./'\._¸ ¸.·¤**¤·.¸.·¤**¤·.. *·. .·* *Andy Dale* /.·*·.\*.·¤**¤·.¸¸.·¤**¤·.* andycool22@peoplepc.com AndyCool22 on AIM ICQ # 76178574 MSN Messenger andycool22@msn.com Yahoo! Messenger - andycool22 ________________________________________________ PeoplePC: It's for people. And it's just smart. http://www.peoplepc.com --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.