What Quantization type are you using?

[jwu42]

I am learning it does depend a bit on the source, but is there a certain combo you think is universally the better than the others listed?

[PowerMacG4]

None of the above. H.263 and H.263.

<edit>
Oh yeah, and you can't use any of the modulated ones without first using H.263. So using MPEG quant in your first pass wont work right.
</edit>

[Alxemi]

What is the difference between (H263 & NewHQ) and (MPEG & NewHQ)?
¿?

[SiXXGuNNZ]

I don't follow, you mean use different ones for each pass?

I normally use MPEG

[mdrobnak]

Originally posted by PowerMacG4
None of the above. H.263 and H.263.

<edit>
Oh yeah, and you can't use any of the modulated ones without first using H.263. So using MPEG quant in your first pass wont work right.
</edit>

Huh? I've been following this advice:

With Koepi’s builds, if you chose “MPEG” as the first pass
“Quantization type”, now you can consider choosing “New Modulated HQ”
Quantization type as many users/testers report to have got pretty good
results with it. If you choose “New Modulated HQ” now for second pass,
H.263 quantization type will be used for quantizers </= 3, and MPEG
quantization type will be used for quantizers >/= 4.


That's from iago's 2 pass XviD guide...It's what I've been using, MPEG 1st pass, New Modulated 2nd pass, and gotten pretty dang good results on my HDTV encodes...

Don't Say A Word and Driven were done at 1120x630 resolution, Driven completely filled a DVD-R with approx. 2MBs to spare, and Don't Say a Word compressed better, I think to about .5 DVD-R -- I probably could have gone to my next resolution:

I have encoded CSI and CSI:Miami at 1280x720, although I kept the first pass and never did pass 2 -- At Quant 2 MPEG Quant, it looked excellent -- and the episodes were only about 2GBs each for about 45 mins! Considering they were about 8GBs when I started, that's VERY good recompression!

So, aside from my good luck with CSI -- Did I do the wrong thing by choosing MPEG as the first pass quant for Driven and Don't Say A Word?

Thanks,
-Matt

[Alxemi]

I don't follow, you mean use different ones for each pass?

Yep, i mean the difference between the options 2 and 4 of the pool. I thought that if you used New Modulated HQ in the second pass, You must use MPEG for the first (I dont remeber where i read it)

[sysKin]

Just remember not to use any modulated quantization with b-frames, the results will be quite unpredicible.

btw what about custom matrices in the pool? :rolleyes:

Radek

[Koepi]

Heya Radek,

look again, there they are ;)

@all:
Btw., use the search function, extend the search to a looong time (not only the last 30 days) and find the thread where we did some research on modulated quantizers.

I.e. if your 2nd pass average quantizer is >3 then you have to use the quant type which gets used for quant >3 in your selected mode (i.e. "old" modulated with mpeg quant type for quants 2+3 and h263 for higher quantizers. if your average quantizer is >3 (i.e 3.01 or 4) you should use h263 quant type for first pass. The quantizer distribution gets more optimal that way)

Regards
Koepi

[bond]

Originally posted by Alxemi
I thought that if you used New Modulated HQ in the second pass, You must use MPEG for the first (I dont remeber where i read it)I also thought that this was the case? So what are the differences between h.263 and Modulated or New Modulated HQ?

Originally posted by sysKin
Just remember not to use any modulated quantization with b-frames, the results will be quite unpredicible.Interesting, I didnt know that...

[jbzzz]

I normally use MPEG-MPEG, one pass-two pass

@jbzzz:)

[jwu42]

Thanks for the replies so far - I have good crisp results with mpeg-mpeg and follow with H.263 and New Modulated as a close second.

[Koepi]

I close this thread now as I gave some hints for scientific backgrounds how to use quantization type modulation.

Such a poll doesn't help finding good results.

Koepi