Is it a good idea to make the first pass at q3 when using 6of9_hvs?

[Chainmax]

I know it is when using standard 6of9, but have no idea about 6of9_hvs.

[Teegedeck]

Yes, it is a good idea to do the first pass of any high-bitrate matrix at quant=3.

Or, more aptly, it is always good to do the first pass at a quantizer that you can realistically expect to reach during second pass at least for a part of the movie. It makes for less quantizer fluctuation.

[Chainmax]

Cool, thanks :).

[Richard Berg]

So if you're aiming to *reduce* bitrate, does that mean you should do the first pass at quant 4 or higher? If so, then you could argue the default 2-pass behavior is wrong, considering the average Xvid user is not going for ultra high quality.

[Sagittaire]

make compressibility test with your matrix and q2:

for ~100% and less: use q2 for first pass with high overflow or [2-31] quant interval ...
for ~75% and less: use q3 for first pass
for ~50% and less: use q4 for first pass
for ~25% and less: use q8 for first pass

make first pass with the same average quant (I and P average quant) than second pass is very good for size quant prediction -> very constant quant for second pass and very better target bitrate (overflow could be low ... 5/5/5 for example)

This methode is in fact a "3 pass mode" with XviD ... lol

[Teegedeck]

Originally posted by Richard Berg
If so, then you could argue the default 2-pass behavior is wrong, considering the average Xvid user is not going for ultra high quality.

It's not that easy, really. I'm more inclined to believe that the average user is under the assumption that he/she should only use quant=3 or lower. Or if he/she is ignorant of quantizers he/she will probably aim for a 2-CD encoding of a 100-minute movie at a 640x resolution. Which again means quants 2-3 (H.263), unfortunately.

On the other hand, I would think it a very good idea if AutoGK chose a fitting quantizer for 1st pass depending on its compressibility check...