Newbie question : LAME 3.xx.xx ?
Reply #44 – 2005-06-01 19:55:57
The current recommended compile is 3.90.3. Recommended compiles are chosen solely on the grounds of proven audio quality (to the best of my knowledge). Therefore, the recommended compile provides greatest audio quality, according to the administration. I can't believe it. It has been explained several time. Reason is also explained on the recommended compile thread you've mentionned. Are you blind? >3.96.1 seemed as though it might be about the same quality as 3.90.3 > in some of the tests done after it was released . The "administration" doesn't claim here that 3.90.3 is better than modern lame release. But Dibrom (or another administrator) followed: > However, in the minds of the administration of this forum it has not yet > been tested extensively enough, thus 3.90.3 is still the recommended version. Isn't it clear enough? Or can't you read? 3.90.3 is recommended because tests were done by some people in the past. And not because quality is proven to be superior to any other lame version . It should be clear enough.The subsequent versions of LAME were optimized for speed. False. Gabriel and Robert worked to improve quality as well as encoded speed. All recent tunings made on recent alphas had no impact on speed but on quality. Read changelog if needed. And if you can't understand it, just count the lines coloured in red...These speed optimizations caused degradation of audio quality, resulting in builds that produced poorer audio quality, and requiring retuning to provide quality equal to 3.90.3. Good Lord? What's this nonesense? LAME is not GOGO... To improve speed and lower quality in the same time, it should be easier to make -q 9 as default. I wonder why developers were irrational enough to spend so many hours to work with various switch in order to achieve the same results.So, the rationale for using versions of LAME produced after 3.90.3 is to have speedier MP3 creation, and the statement about the tradeoff between quality and speed was created. For someone who can't ABX difference between 2 release of lame, you have many things to tell...That is how I reached that conclusion. Did you read HA TOS? Conclusions and claims must be based on listening tests, and not on syllogism or phantasmas. All modern encoders were able to gain speed and quality in the same time: Vorbis, MPC, AAC [Nero, faac and QT]. And nobody on this board conclude that higher speed necessary lead to lower the quality. Except maybe some people with fertile imagination but no free time to contribute to help on encoder's development.I've acknowledged that I am unable to ABX at the bitrates given and am going solely on the opinions of others. Then read all tests that were performed on lame during the last year. You'll find several evidence that modern release of LAME are all except "suboptimal" compared to 3.90.3. I've spent several hours (probably more than 100) to test lame, and I've published most results on this board. Some other people really interested about quality and progress also spent some of their free time to give positive or negative feedback. You should better do the same, rather than polluting this board with unfounded equations about quality and biased interpretation of HA.org recommendation. And if you really don't want or simply can't work to improve LAME, read at least the informative tests that are posted on this board: you'll probably learn interesting things and therefore stop spreading wrong informations to other people. HA.org as well as LAME developers need testers and feedback, not parrots and imagination.