For WMA9 VBR, does 2 pass provide a quality increase over 1 pass (assuming same file size)? Or is the extra pass simply for being able to produce a VBR file with predictable size (i.e. ABR), with no increase in quality over 1 pass?

VBR is actually more like ABR. It uses the two passes to make it the VBR setting you choose. Like 128k VBR. Quality VBR is true VBR and cannot have 2 passes. So apart from Quality VBR the other VBR are actually ABR. Now CBR does offer 2 pass encoding i don't know what for though.

2-pass encoding generally has two benefits: more accurate filesize prediction, and improved quality due to the ability to view frames of audio/video collectively, rather than in a linear frame-to-frame manner. The latter benefit allows the encoder to view sections of frames which require more bits, and sections with less need, and allocate bits appropriately. Does this answer your question?

edit: concerning 2-pass CBR, I have no idea what it's for, maybe they manipulate the bit reservoir somehow to get higher quality.

2-pass CBR influences both bit reservoir and block switching. At least that's what one of the MS devs told me in the newsgroups sometime ago when I asked for it.

So would 2 pass VBR 192 offer superior quality to 1 pass Quality 98 (which produces avg. bit rates of 275-350)?

I'm assuming the latter is better, but I am not an expert.

If I understand your question properly, then you might be comparing apples and oranges.

Let me explain: when you encode something with two-pass, you almost always give the encoder a nominal average bitrate, which it tries to achieve by allocating bits appropriately over the span of the material. The encoder's programmers may also take the first pass's information to improve general quality by various means. When you're encoding with a quality setting (one-pass), then instead of modifying quality to suit space (and thereby getting an assigned average bitrate), you instead forfeit predicable file size for constant quality.

For instance, let's say that you wanted to encode a song. In two-pass mode, you'd be able to assign it an average bitrate. However, while you are garaunteed the highest quality for the given average bitrate, the encoder still may be able to give you better quality. It would be incapable of doing so though because of the file size limitations which you imposed. If you encoded the file at 100% quality, then the encoder would do it's absolute best to achieve the greatest quality possible for every frame, disregarding file size completely.

If this still makes sense to you, then you can see why both modes are benefitial. If you need a constant quality, then choose the quality setting. If you need a predictable file size, choose 2-pass. Does that all make sense?



edit: don't assume, of course, that higher bitrates mean higher quality. That's a pretty popular myth, actually. The important thing to remember is that for the material you're encoding, there will be a peak quality that the codec you're using can achieve, and beyond that you're most likely going to be wasting bits. If WMA can achieve top-notch quality at 192kbps 2-pass, then it would be wasteful to force it to use more bits for sound you're most likely unable to hear anyway.

That is exactly what I thought but its great to have it confirmed by an expert.



If WMA did not need the bits to create a capture the sound, why would it use them? Also, any idea why WMA 98 uses bit rates so much higher than LAME APS or even LAME API?

Typically, this comes from two (or more) reasons. First off, the encoder may be fooled into "thinking" that it needs more bits to properly mask the sound of an encoded sample, and therefore inflate the bitrate. Also, not all encoders are properly optimized. But hey, being WMA I'd think it would be as aggressive as possible. Of course, when you're comparing 1- and 2-pass, you have to understand that even though 192 may be the *average* bitrate, it may well climb much higher than that to achieve quality on troublesome frames. Therefore, you can see why 192kbps ABR 2-pass would be able to achieve as high a quality as higher-bitrate one-pass encodings. Basically, the one-pass encodings can't look back and save/redistribute bits as the 2-pass mode can, and so it has to simply throw bits at the problem. 2-pass can at many times achieve similar or even greater quality at lower bitrates, simply through quality optimizations that are performed on the second pass.



This may be caused by the same reasons I described above. LAME may well be tuned to cut out more info than WMA is, and therefore achieves lower bitrates. I'm not an expert, though, and I'd love some input from people that are.

Great explanations! Thanks. So 2pass VBR's main purpose is to obtain a targeted bit rate, but the 2nd pass can also be used to achieve further quality improvements if this feature has been implemented in the encoder. Is it known whether the WMA9 encoder takes advantage of the 2nd pass for more than just bit allocation to achieve a targeted average bit rate?

In other words, if the same song was encoded twice using WMA9 (once in 1pass VBR and once in 2pass VBR) and coincidentally happened to have the exact same average bitrate, would the 2pass encode would be of higher quality?

Theoretically, yes, because short transients and other tough-to-encode sounds would be predicted, and prepared for, whereas the 1-pass encoding would be forced to accomodate in other ways (i.e., increasing the bitrate). And hey, I could be wrong about this would anyone else like to comment?