DivX® 6.1 codec Patch 2 beta

[DigitAl56K]

Hey guys,

I just posted an update to DivX 6.1 codec (http://labs.divx.com/archives/000068.html) up on DivX Labs. This should fix several issues, including the crash when selecting psychovisual mode that some of you reported.

You'll also find a ton of other good stuff has been posted up to DivX Labs (http://www.divxlabs.com) over the past week or so, including Converter Beta 2 (http://labs.divx.com/archives/000071.html), and footage from CES (http://labs.divx.com/?q=CES06) shot by our very own team! In front of the camera you see Ben, also known on the forums as "thekid". You can download him in standard or high definition (you know you want to see him in HD..), and you can also watch the video in your browser using the new beta DivX Browser Plug-in (http://labs.divx.com/archives/000072.html).

[mod]

1st positive feedback: it works :)

Athlon XP 2700 Thoroughbred

[Romario]

2st positive feedback: it works very good

Athlon XP 3200+ Barton core

[DigitAl56K]

Patch 2 now up, which really fixes all of the issues! My bad ;)

Have a good weekend folks!

[CruNcher]

@DigitAl56K
The CES Coverage is cool also the new Website Design is great but i never gonna understand why it isn't centered :P

[akiza]

Does it means DIVX.COM will release CODEC 6.11 ?..^_^

If it ture, when we can get the new codec ?..

[DigitAl56K]

Yes we'll also be releasing an update in the bundles on DivX.com once the patch on DivX Labs has been more widely tested :) You can help by letting us know how it worked for you either here or over at forums.divx.com - we read both frequently.

[Slitheen]

Before I download converter beta2 does it do two passes? Not really a bug as such, but I found the single pass quality even with high bitrates not very good because of it.

[Sagittaire]

Very better sensivity for bframe (not only letter box fix I think). In practice DivX 6.1.0 don't use bframe ... ???

[Isochroma]

If someone already installed patch 1, can they install patch 2 on top of it, or must they do a complete uninstall/reinstall first?

[DigitAl56K]

Very better sensivity for bframe (not only letter box fix I think). In practice DivX 6.1.0 don't use bframe ... ???

There was a problem in 6.1 where some letterboxed content would not have b-frames. If you had cropped the borders you wouldn't experience this effect.

If someone already installed patch 1, can they install patch 2 on top of it, or must they do a complete uninstall/reinstall first?

You just need to install patch 2 over the top :)

[Sagittaire]

There was a problem in 6.1 where some letterboxed content would not have b-frames. If you had cropped the borders you wouldn't experience this effect.


- More bframes for b2 in all situation (I test with or without crop, 240p/400p/720p resolution, ~q8/q4/q2 average quant multipass)

DivX 6.1.0 b2
I-VOPs: 46 (2,23%)
P-VOPs: 1291 (62,61%)
B-VOPs: 725 (35,16%)
S-VOPs: 0 (0,00%)
N-VOPs: 0 (0,00%)

Max consecutive B-VOPs: 3
1 consec: 70,78%
2 consec: 20,87%
3 consec: 8,35%

DivX 6.1.0
I-VOPs: 46 (2,23%)
P-VOPs: 1612 (78,18%)
B-VOPs: 404 (19,59%)
S-VOPs: 0 (0,00%)
N-VOPs: 0 (0,00%)

Max consecutive B-VOPs: 3
1 consec: 74,68%
2 consec: 19,48%
3 consec: 5,84%

- IMO Big problem for RC : for q2 average encoding there are sometime frame with quant 20. I use unconstrained mode with Bitrate Modulation = 0. With exactly the same bitrate constant quant mode is very better for OPSNR. Constant quant encoding is visually very better than multipass encoding.

[DigitAl56K]

- IMO Big problem for RC : for q2 average encoding there are sometime frame with quant 20. I use unconstrained mode with Bitrate Modulation = 0. With exactly the same bitrate constant quant mode is very better for OPSNR. Constant quant encoding is visually very better than multipass encoding.

Can you please provide the CLI settings you used for each pass, as well as some details of the source material.

I've ran a couple of tests here at home and I don't see this problem.

[Sagittaire]

Can you please provide the CLI settings you used for each pass, as well as some details of the source material.

I've ran a couple of tests here at home and I don't see this problem.

Certainely ...

first pass
-bvn1 6000000 -vbv 27336400,100663296,75497472 -dir "C:\Program Files\DivX\DivX" -w -pq 8610 -b 3 -quantization=2 -enable_feedback=1

second pass and third pass
-bvnn 6000000 -vbv 27336400,100663296,75497472 -dir "C:\Program Files\DivX\DivX" -w -pq 8610 -b 3 -quantization=2 -enable_feedback=1

sample is here (Iframe quant 23 here)
http://multimediacom.free.fr/Video/Bug-RC.avi

I have sample with multiple probleme for multiple consecutive Iframe. Seem to be a big problem only for HD source ... OPSNR don't like high variation for quantizer and my eyes too ... ;-)

[Isochroma]

I use constant quant encoding as well (q=1) so this is definitely something that ought to be checked...

[akiza]

Yes we'll also be releasing an update in the bundles on DivX.com once the patch on DivX Labs has been more widely tested :) You can help by letting us know how it worked for you either here or over at forums.divx.com - we read both frequently.

Is there any plan to add some optimize for sse3/lddqu in the future ? :D

[fight2win]

divx encoding was crashing on my pc, i'll try this patch, i hope it works now!

[Romario]

Is there any plan to add some optimize for sse3/lddqu in the future ? :D

Yeah, I also ask same question.

[acrespo]

- More bframes for b2 in all situation (I test with or without crop, 240p/400p/720p resolution, ~q8/q4/q2 average quant multipass)

DivX 6.1.0 b2
I-VOPs: 46 (2,23%)
P-VOPs: 1291 (62,61%)
B-VOPs: 725 (35,16%)
S-VOPs: 0 (0,00%)
N-VOPs: 0 (0,00%)

Max consecutive B-VOPs: 3
1 consec: 70,78%
2 consec: 20,87%
3 consec: 8,35%



How can I see these numbers (I,P,B,S and N frames)? I tried a tool named DRF Analizer but this tool only report P-Frames in my video, but the feedback window show me about 25% of B-Frames too.

Andre Crespo

[iNFO-DVD]

MPEG4 Modifier

[Sagittaire]

1) Always very big problem for Rate Control : too high variability for quantizer -> very bad OPSNR in comparison with XviD for low average quantizer encoding ... ???

2) Bframe sensivity is strange:

test with 1280*720 6000 Kbps King Kong Trailer full screen without letter box.

DivX 6.1.0 patch 2
Packed bitstream: Yes
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263

I-VOPs: 172 (4,78%)
P-VOPs: 3268 (90,83%)
B-VOPs: 158 (4,39%)
S-VOPs: 0 (0,00%)
N-VOPs: 0 (0,00%)

Max consecutive B-VOPs: 3
1 consec: 88,49%
2 consec: 9,35%
3 consec: 2,16%


DivX 6.1.0
Packed bitstream: No
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263

I-VOPs: 176 (4,89%)
P-VOPs: 3424 (95,11%)
B-VOPs: 0 (0,00%)
S-VOPs: 0 (0,00%)


XviD 1.2
Packed bitstream: Yes
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263

I-VOPs: 114 (3,17%)
P-VOPs: 1525 (42,36%)
B-VOPs: 1960 (54,44%)
S-VOPs: 0 (0,00%)
N-VOPs: 1 (0,03%)

Max consecutive B-VOPs: 2
1 consec: 41,30%
2 consec: 58,70%
N-VOPs: 0 (0,00%)


test with 1280*720 3000 Kbps King Kong Trailer full screen without letter box.

DivX 6.1.0 patch 2
Packed bitstream: Yes
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263

I-VOPs: 133 (3,70%)
P-VOPs: 2624 (72,93%)
B-VOPs: 841 (23,37%)
S-VOPs: 0 (0,00%)
N-VOPs: 0 (0,00%)

Max consecutive B-VOPs: 3
1 consec: 91,78%
2 consec: 6,66%
3 consec: 1,57%


DivX 6.1.0
Packed bitstream: Yes
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263

I-VOPs: 134 (3,72%)
P-VOPs: 2913 (80,96%)
B-VOPs: 551 (15,31%)
S-VOPs: 0 (0,00%)
N-VOPs: 0 (0,00%)

Max consecutive B-VOPs: 3
1 consec: 95,24%
2 consec: 4,57%
3 consec: 0,19%


XviD 1.2
Packed bitstream: Yes
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263

I-VOPs: 116 (3,22%)
P-VOPs: 1538 (42,72%)
B-VOPs: 1945 (54,03%)
S-VOPs: 0 (0,00%)
N-VOPs: 1 (0,03%)

Max consecutive B-VOPs: 2
1 consec: 43,90%
2 consec: 56,10%

Encoder seem really good (constant quant mode) and better than XviD for metric but low sensivity for bframe and strange RC are a big problem for metric and visual quality in multipass mode

[DigitAl56K]

Thanks Sagittaire :) I'll work with out team to check into this one today.

[Selur]

Psychovisual Bug seems to be fixed here too :D

[acrespo]

Yes, I also confirmed that psychovisual bug is gone.

The quality of video up with Patch 2 because B-Frames is correct now. I have 25-30% B-Frames (0% in version 6.1).