Why is Default Q range 1 to 31 now?

[Zep]

Why is Default Q range 1 to 31 now?


Is there a reason it was changed from 2 to 31?



Thx


Zep

[crusty]

Reason was simple:
No more undersized files. If the filesize turns out to be to small, quant 1 is used. (if you like you can still use quantizers in range of 2-31, quant 1 just eats bytes without looking that much better).

[Dali Lama]

"quant 1 just eats bytes without looking that much better."

Hi Crusty, in my experience, Quant 1 does look significantly better in extremely difficult to encode scenes (ie: dark, low contrast, high motion, noisy, etc.). Yes, it does increase bitrate a lot too.

Sincerely,

Dali

[Gazza]

I was under the impression that including quant 1 resulted in problems with meeting final filesize targets. I stopped using quant 1 ages ago because of this.

[crusty]

in my experience, Quant 1 does look significantly better in extremely difficult to encode scenes (ie: dark, low contrast, high motion, noisy, etc.).
I see, like really difficult stuff like fog, steam, smoke etc.

I was under the impression that including quant 1 resulted in problems with meeting final filesize targets. I stopped using quant 1 ages ago because of this.
RC1 had this issue. RC2 and RC3 have this fixed. For earlier betas, I don't know. For alphas it was not possible to use Q1, it would just reset to 2.

[Gazza]

Thanks Crusty. I will give it a go again. I have always been impressed with the ability to hit filesize when using 2-31 quants. It will be interesting to test the latest xvids

Gazza

[gatormac]

What I have had success with is setting my Q's at 2 to 31 and adjusting difficult scenes like smoke and fog or extremely dark scenes with a zone set to Q of 1. At least it seems to have worked for me.

[alexnoe]

RC2 definitely did not have it fixed. The movie in question was DS9 6x02 where i tried it.

[Koepi]

alex:

you're sure you didn't import "old defaults" somehow?

We have some "smart" short-time increments of the overflow treatment vlaues in case of q1 frames (dunno, can be that this is a post-RC2 feature, but i think it's older).

If you didn't play around with the values (i.e. quant range from 1-2 :P ) then please file a proper bugreport with all your settings please.

Koepi

[alexnoe]

I just run an encoding this night with RC3 and Quantizer 1-5, and it worked perfectly, so no need for a bug report :-)