Xvid quantizer stuck at 2.00

[helix]

I've been having this problem for a while now, It's started around when the fourms went down of cource, and I just can't seem to fix it. I've tried everything from reinstlling Xvid to reinstlling just about every codec I have. I can't get Xvid's 2nd pass quantizers to go past 2.00. The min and max quantizer for all the frames types are clearly 2-31 in the options, but I still get the following picture on the second pass. I've also confirmed this with Divx analyzer and it comes out to be 2.000 in every encode I do.

http://www.clanbbw.com/publicimg/data/images/quantizers.jpg

[Manao]

You are saturating the codec : you try to reach a filesize which is larger than the size the encode would get at quant 2. So XviD put only quant 2 frames. If you want other quants to appear, lower the filesize you're aiming at.

For more details on the behavior, make some searches here on "saturate" or "undersizing"

[helix]

I was aiming for 233Mb for 23 minutes which didn't seem like it was too big a file size at all. I tried going down to 175Mb for 23 minutes and i'm still getting a 2.00 quantizer on the 2nd pass. :confused:

[Manao]

Did your two attempts reach the aimed filesize or not ? I'd bet it didn't. If it reached XXX MB, then ask for 2/3 of XXX MB, and then you'll get other quantizers than 2.

[helix]

yeah, I did get my desired file size. I wanted 7330Kb of video and 814Kb of audio for 1 minute and my end file size was 8194Kb. The quantizer for clip was 2.044. What i'm also wondering is why it says "2.00" under the max quatizer during the encode, when the max is specified as being 31.

[shutdown]

What i'm also wondering is why it says "2.00" under the max quatizer during the encode, when the max is specified as being 31.

That is the real-time stats... it means that the encoder hadn't reach more than Q2, if it will (even for a single frame) you'll see that in the "Max".

[superdump]

Check you haven't forced quant 2 in a zone.

EDIT: I misunderstood. I see you are hitting your filesizes correctly, right? It's just the statistics display that isn't correct.

[helix]

I've been checking the actual files though, and the quantizers are really all 2.00, so the display must be correct. It always used to say "31" insted of "2" for the max. This problem started almost right away like I did somthing in the settings by mistake, everything looks good though. It's most confusing and frustrating, I have a bunch of stuff lined up I have to encode right now too.

[Stian]

It always used to say "31" insted of "2" for the max.

Unless you're encoding at super low bitrates or have specified a low bitrate/quat zone somewhere, the real time stats should NEVER show a max of 31!!

[Leak]

Originally posted by helix
I've been checking the actual files though, and the quantizers are really all 2.00, so the display must be correct. It always used to say "31" insted of "2" for the max. This problem started almost right away like I did somthing in the settings by mistake, everything looks good though. It's most confuisng and frustrating, I have a bunch of stuff lined up I have to encode right now too.

The "Min" and "Max" values in XviD's status window just show what the minimum and maximum quantizers that were really used in the encoded frames are (and it's been like that as far as I can remember). If your source is so compressible that the filesize will not or just barely be reached using all Q2 frames, Min and Max will both be 2, so where's the problem?

This just means your encoded frames have the best quality possible (short of using Q1), so why do you even think that it's a problem?

I've encoded the first volume of Witch Hunter Robin at 640x480, and all but the 4th episode were so compressible that the files were all Q2 and up to 50 MB below the target size - which means when I'm going to encode all 26 episodes I'll have more spare room for extras and other less compressible episodes when I want to fit everything on a DVD+R...

If you really want to make your source hit the filesize, up the resolution, sharpen the image or use a custom matrix, but you just can't add details that aren't in the source that way...

np: Company Flow - Krazy Kings (Funcrusher Plus)

[Trahald]

Its very easy to test the theory.. just go for some rediculously small size like 10mb.. if it hits q2 for 23 mins worth of footage at 10 mb then something else is afoot (assuming your resolution is at least 320xXXX)...

[Leak]

Originally posted by Trahald
Its very easy to test the theory.. just go for some rediculously small size like 10mb.. if it hits q2 for 23 mins worth of footage at 10 mb then something else is afoot (assuming your resolution is at least 320xXXX)...

True. Or just use white noise as input and go for the target size you originally wanted... :p

np: B.Fleischmann - Take Your Time (Welcome Tourist)

[Didée]

Uh, pure white noise is a bitch to encode. One should at least apply PixieDust() on that. But this will take away the sharpness, so it should be followed by 4x SSXSharpen().

Or alternatively, feed the noise into iiP(), since this will do everything together.

/* runs away as quickly as he can */

[Leak]

Originally posted by Didée
Uh, pure white noise is a bitch to encode.

Well, that's the whole point - I hope nobody wants to actually watch 25 minutes of white noise... ;)

I guess the filters you suggest rather make it gray noise, or something... :p

/* runs away as quickly as he can */

Oooh no you won't! *whack*

np: B.Fleischmann - Take Your Time (Welcome Tourist)

[helix]

Well I did do a little test of sorts. I put it down to 140Mb an episode (23 min) and got a average quantizer of 3.1 so it looks like that was my problem. I've never encoded anime that can compress that well, especially with all P-frames, which is why I was sceptical I guess... I feel kind of stupid now. But anyway, thanks alot for the help guys.