2-pass "target size" vs 1-pass "target quant"

[gamaD]

I read that if size exact size isn't an issue, 1-pass Q2 is more or less the way to go.

But just to make things clear quality-wise.
If I encode a movie 2 ways:
1) one pass Q2, ending up with a 1 200 000 000 bytes filesize,
2) two pass, entering 1 200 000 000 bytes as target size,
with all other settings the same, do I end up with the SAME quality? (ignore the possibility for two pass encode to be a slightly different filesize)

[lamer_de]

If you use q2, probably yes. q2 is regarded as maximum quality (q1 were introduced tocompensate for too small encodes. Some people have reported they deliver better quality than q2 frames, but I never checked that, so I can't confirm or deny). So, the 2 pass process usually shovels around quantisizer based on the statistics it gains in the first pass. If there is no distribution of different quants needed, because the codec is always saturated, there's little todo for the 2nd pass. Maybe it'd use some q1 frames and some q3 frames, but that's just guessing from my side. 2pass is only usefull if you want to achieve good quality on a certain filesize. As you don't care for filesize when you encode with fixed quants, 2pass should offer no benefits for you. A different szenario would be if you would encode at, let's say quant 4. I could imagine a 2 pass encode for the same filesize would do look better in specific scenes. But as I never encode with fixed quant, I can't say 100%

CU,
lamer_de

[pelle412]

I think the answer to your question is Yes. Several tests conclude that with a high enough bitrate using 2-pass, the quantizer distribution becomes the same as 1-pass Q2.

Now to another question related to this. Which encoder options does not make sense for 1-pass Q2 encoding? Here are some which I have not been able to conclude that they add value:

Adaptive Quantization
Trellis

Can some oracle comment on the value and/or function of these features in a 1-pass Q2 type encode?

[sysKin]

Originally posted by pelle412
I think the answer to your question is Yes. Several tests conclude that with a high enough bitrate using 2-pass, the quantizer distribution becomes the same as 1-pass Q2.

Now to another question related to this. Which encoder options does not make sense for 1-pass Q2 encoding? Here are some which I have not been able to conclude that they add value:

Adaptive Quantization
Trellis

Can some oracle comment on the value and/or function of these features in a 1-pass Q2 type encode?
Adaptive quantization simply means that your file will not longer be constant quant. Some parts of some frames will get higher quant - and file will be smaller, while should (if lumimasking was good) look the same.

Trellis - like always - means that quality will be better, file will be smaller, and encoding will be longer. Same as for VHQ, chroma motion, or higher search precision.

Radek

PS. As for original question, single-pass gives in theory better quality. This is because the quality is trully constant. Second pass always fluctuates quality a bit to accomodate its bitrate curve, and this never matches real video perfectly - so the quality is not perfectly constant.
Single pass is also faster, of course ;)