DivX Manihi Codec Beta

[Gej]

DivX Manihi Codec Beta is up at DivX.com

http://www.divx.com/divx/windows/beta/


Welcome to the new codec beta release. The release is full of new features that will improve the overall quality of your digital video experience. A list of new features and fixes can be found in the "What's New" section below. The Quality Assurance (QA) period will run for about three weeks starting 2003-07-25. All the results must be e-mailed back to us by 2003-08-15 at the latest. Intermediate results are welcome. Please submit your bugs through our bug submission form .You can discuss your bugs and feedback with the DivX community in the DivX Beta Release Feedback Forum.


What's New:

* High Quality encoding mode
- Different motion estimation criteria
- Rate distortion algorithm

*High Quality psychovisual mode
- Texture cortex masking and Rate distortion algorithm
- Hybrid mode (Texture cortex masking)

* High Quality B frame mode
- The motion estimation for B frame has been improved
- Quantization modulation is changed

* Feedback mode
- Allows visual feedback during encoding

* Decoder rearchitecture
- Decoder is 10 to 30% faster on all CPU platforms
- Decoder supports auto-postprocessing

* Encoder GUI
- The codec GUI has been redesigned to ease accessibility and simplify overall usage.
- RC modes: Enabled complexity vector logfile, and experimental texture / motion modulation switch
- RC to support real time control through API support
- I frame insertion forced through API support


Fixed:

* Fix for quantizer overshoot when over bitrate following a transition to a complex scene
* Change in quantizer from one frame to next limited to change in ideal quantizer or +-2
* Fix for a bug in RGB32 to YUV conversion on P4
* A minor issue with Multiple CPU configuration
* Fix for broken B-frames ('P' in FrameSpec was occasionally producing an I frame)
* 3.11 Compatibility improved (NO mismatach at all on ALL the clips tested, more than 60 hours of contents!)
* RC VBV enforce "3 sec" max bitrate
Bugfixes, performance improvements in nth-pass strategy planning
* MV reuse is functional
* Qpel, B frames and slow and slowest mode doesn't generate artifacts anymore

[cordraconis]

d**n, I was looking around on your website becouse I have some Probs with Dr.Divx, and I just saw your new béta.

It seems you were faster than me to report it to Doom9, héhé.

:p

[SiXXGuNNZ]

very nice!

when I get my new dvd-rom drive next week I will be testing this thing out :)

well I tested it right quick on some avi sources, but that isn't very accurate :D

[The People's Elbow]

In VDubs compression dialogue it still says "kauehi" though codec seems to have changed - at least the settings did.
Is it me or a little bug in the frontend?

greetz Elbow!

[SiXXGuNNZ]

Originally posted by The People's Elbow
In VDubs compression dialogue it still says "kauehi" though codec seems to have changed - at least the settings did.
Is it me or a little bug in the frontend?

greetz Elbow!

I dunno ay, mine says DivX Pro Manihi Codec

[Animaniac]

Originally posted by The People's Elbow
In VDubs compression dialogue it still says "kauehi" though codec seems to have changed - at least the settings did.
Is it me or a little bug in the frontend?

greetz Elbow!

I needed to uninstall Kauehi before installing Manihi, for it to work correctly.

[rozemab]

Is it possible to install Manihi without messing up my Divx Pro install?

I would like to try it out with vdub etc. and not affect my Dr Divx configuration.

[Gej]

You can install Manihi safely, it'll not impact DrDivX, as it use it's own private encoder.

[saeed]

I have just downloaded and installed manihi. but what I got was kauehi in the decoder and the encoder. The previous version on my system was 5.03 not kauehi. the reamde file is for manihi, though.

The file size is 1,912,749 bytes. Can someone confirm that.

[The People's Elbow]

I had to delete the divx.dll manually to get manihi working, maybe my system is messed up a bit ;)
This new overlay dialog is a pretty thing, but for job encoding it's a disadvantage, because its always showing the encoded output which slows down encoding while I'm asleep... can that be disabled?

greetz elbow!

[pluonk]

I installed Manihi today and am now encoding a movie with VDubmod (which btw correctly identifies as "DivX Pro Manihi Codec"). It runs extremely slowly, ~7fps, while 5.05 and previous encoded at 15-20fps (for the same type of movie, which is PAL interlaced using FieldDeinterlace). The settings for Manihi are the defaults (perfomance/quality: standard, psychovisual enh: slow, -Qpel, +GMC, +Bframes). Has anybody tested it or know what's wrong?

Btw, I closed the overlay dialog, hoping that it was to blame for the slow performance, but it didn't help at all. If I choose to keep this codec, does anybody know how to prevent this dialog from popping up each time VDubmod runs?

[saeed]

To disable the feedback window you have to click the settings bottun next to OK bottun and check the option.

I guess it is slower because of the new psychovisual enhancement and the motion estimation in B-frames.

[SeeMoreDigital]

Well here we are again with another DivX test codec. Great fun!

To disable the feedback window you have to click the settings bottun next to OK bottun and check the option.
I guess it is slower because of the new psychovisual enhancement and the motion estimation in B-frames.

Granted the decoder will be slower if psy etc is used. But if you disable the feedback window the encoder should run approx twice as fast.

I'm currently encoding Tomb Raider 1 (PAL 96min) in anamorphic (720x576) with video at 913kbps and audio at 96kbps. So I should have some results in a few hours.

As per normal I'll compare this new encode against previous encode versions such as DivXPro5.0.5 & Kauehi. And view the output on my TV via the Xcard.

[Fr4nz]

Originally posted by pluonk
I installed Manihi today and am now encoding a movie with VDubmod (which btw correctly identifies as "DivX Pro Manihi Codec"). It runs extremely slowly, ~7fps, while 5.05 and previous encoded at 15-20fps (for the same type of movie, which is PAL interlaced using FieldDeinterlace). The settings for Manihi are the defaults (perfomance/quality: standard, psychovisual enh: slow, -Qpel, +GMC, +Bframes). Has anybody tested it or know what's wrong?

Btw, I closed the overlay dialog, hoping that it was to blame for the slow performance, but it didn't help at all. If I choose to keep this codec, does anybody know how to prevent this dialog from popping up each time VDubmod runs?

Also the GMC active slows down the encoding ALOT.
Try to put psyco on fast and disable GMC..it should be really faster.

[SeeMoreDigital]

My first test with DivX Manihi is complete.

To be honest I can't see any difference between this version and Kauehi. I will generate a lower bitrate encode tomorrow. Maybe (I hope) this will reveal some more positive results.

All is not lost though as I do prefer the new interface.

Anybody else like this?

[Fr4nz]

I also tried Manihi at 1000kbps and I can't see any difference with Kauehi (same options enabled). Oh I have to say also that I find Kauehi faster, so I think I'll stick with it.

[Morbo]

Overlay window is ultra cool:cool: :cool:

Looks like a nice release overall again gej..

Cheers!!

[silver_cpu]

I have an interesting problem...

Ever since I upgraded (from Kauehi), whenever I try to encode through VirtualDubMod , it gives me this error:

VideoSourceAVI error: the source image format is not acceptable. (error code -2)

Anyone know what this means? I've tried different settings, as well as changing the resolution (thought I might have hit a bad one), but nothing seems to help. I'm trying to encode one of the pieces in "the animatrix," can anyone describe what VDM's problem is, and how to fix it?

EDIT: Well, it seems that this is a system-wide problem. I tried out Xvid and Divx3 just in case it was truely DivX5 that had caused the problem, that didn't fix it. Vor some reason, VDubMod is rejecting the video stream. I've tried uninstalling and reinstalling the GKnot rip-pack (0.28.5, with 0.28.5.2 patch), and un-installing DivX (even running a registry cleaner to catch stray reg entries), reinstalling with the retail 5.0.5 codec (none of the betas). Nothing works :( It keeps giving me the same error, but I can't seem to locate that error when I search the boards... Maybe I'll have luck and figure out what it is soon.

[LordIntruder]

Look here, it may solve your problem:

http://forum.doom9.org/showthread.php?s=&threadid=51906

[silver_cpu]

Thank you so much! It worked perfectly :)

[LordIntruder]

I've just given a short try to Manihi.

I can compare the speed as I have just encoded a full movie with Kauehi and I use the SAME avs.

With the same parameters, if I select the new psy in slow mode + standard encoding, it's around one hour longer compared to Kauehi new psy + standard encoding.

Now if I select new psy in fast mode, it's almost two times shorter to encode the full movie!?!

However I stopped the tests 10 min after launching them with VDM. By the way encoding time are given by VDM and I can say they are rather accurate. But I will do full encoding tests with the same movie to compare.

I really wonder what are the differences between those two modes quality speaking because the speed improvement in new psy fast mode is awesome. I wonder if it worths to encode in new psy slow if it takes up to two times more time. Some tests have to be done in this area.

Could Gej or any other divx people team tell us, if possible, what are the differences we could expect from new psy fast and slow modes?

Of course if we combine new psy slow with slow encoding it's even slower than Kauehi new psy + slow encoding.

Finally I noticed with Manihi than whether you select new psy in fast mode or you disable it, there is no noticeable improvement in encoding speed. At least with my 1800+ and on a 10min test. So I think the new psy fast mode can always been ticked and now the question is rather or not new psy in slow mode improve the overall image quality or not.

MV has done is come back but I wonder if enable it can lower quality by a bit or not? With older divx 5, 5.01, etc... many people noticed a small loss in quality with MV enabled. Could it be same here? Another thing to look at.

And cheers for the new GUI and the feedback window. It's complicated to analyse but at least I can shine my friends on. :D

[dTb]

Not much to really report just yet but the mvinfo.bin file wasn't generated on my first encode so I'm leaving it unchecked at the moment. Other users have reported this on divx.com.

@SeeMoreDigital, I can't give any figures but disabling the feedback overlay doesn't seem to make much difference. Certainly not double the speed in any case.

I'm unsure atm what the difference is between the fast and slow pve, I suspect fast may be the old method and slow is the new. I'll try to confirm later but if someone wants to check themselves I think I know a way. The old method often generated frames with average quants that were not whole numbers (ie. 4.75) while the new method in Kauehi always generated quants with whole numbers.
Try encoding one file with fast and one with slow and use the osd in ffdshow to check the average quants of frames during playback.

DigitAI56k from divx.com made a comparison and mentioned that encoding speed greatly increased when using q-pel when compared to the standard half-pel. This is the opposite to what you'd normally expect so maybe it's some kind of bug.

I generally like the new encoder gui but having the b-frame option in the profile wizard is a bit of an annoyance (to a lesser extent with gmc and q-pel because I usually don't use them) but is understandable.

In Kauehi the sct was ignored when using the slow and slowest performance/quality settings, I figure this to also be the case in Manihi but am unsure till someone from DARC confirms it. More info can be found on why this is the case in the divx labs forum. Personally I like this and feel it's possibly a reason why high motion is improved in Kauehi.

Just some info I thought people here might find usefull:)

[DigitAl56K]

Please see this regarding the speed of Manihi:

http://forums.divx.com/viewtopic.php?topic=52380&forum=23

And here are some test results I've compiled so far:

http://www.btinternet.com/~digital56k/DX_MB1_R1_D56K.pdf

[SeeMoreDigital]

Like DigitAl56K I've managed to generate some more DivX Manihi tests. But this time I selected a lower bitrate.

My encodes were also generated with the 'Feedback Window' disabled and enabled. And with the 'Show Picture' output disabled and enabled.

According to Gej there should'nt be much difference with the encoding speed but my results don't agree!

Now, for all those people who don't already know. I prefer to encode my output files in anamorphic (720x576) ie with no cropping and resizing and also view the output via Sigma Xcard.


My encode information is as follows: -

Source 'Video Only' File Information
Selected Movie: StarWars 2 PAL
Selected Chapter: 41
Clip RunTime: 06m 17s (377.24 secs)
Total Frames: 9431
Pixel Size: 720x576
Video Type: Mpeg2 (M2V)
Field Type: Progressive
Output AR: 16:9 (1.77:1)
Image AR: 2.35:1
File Size: 243,030KB (237.33MB)

------------------------

Encoded 'Video Only' File Information
Application Used: MpegMediator 1.5
Output Pixel Size: 720x576 Anamorphic
Codec Used DivX Manihi
Bitrate Used 627kbps
Number of Passes 2

------------------------

Encoded File Number 01
Performance/Quality: Standard
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 80%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 768 frames
Selected Profile: Home Theater
Bidirectional Encoding: Unchecked
1st Pass Encode Speed: 07Min 52Sec
2nd Pass Encode Speed: 07Min 54Sec
File Size: 29,105KB (28.42MB)

------------------------

Encoded File Number 02
Feedback Window: Enabled
Show Picture: Enabled
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 80%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 768 frames
Selected Profile: Home Theater
Bidirectional Encoding: Unchecked
1st Pass Encode Speed: 14Min 14Sec
2nd Pass Encode Speed: 15Min 07Sec
File Size: 29,105KB (28.42MB)

------------------------

Encoded File Number 03
Feedback Window: Enabled
Show Picture: Disabled
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 80%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 768 frames
Selected Profile: Home Theater
Bidirectional Encoding: Unchecked
1st Pass Encode Speed: 10Min 09Sec
2nd Pass Encode Speed: 10Min 02Sec
File Size: 29,105KB (28.42MB)

------------------------

In conclusion. Using this bitrate and settings DivX Manihi does not look as good as WMV9 VCM.

Sorry!

But not to be daunted, i'm currently performing some more tests using DigitAl56K settings.

Cheers

[LordIntruder]

DigitAl56K: about your first link I wonder what you mean in:

"Also remember that the encoder isn't configured correctly when you first load it, click the "Restore defaults" button to set it up correctly the first time"

Manihi isn't correctly set up? What is wrong with it? The only thing I notice strange could be the "Scene change threshold: 80%". I would have set it up to 50% but I may be wrong.

Nice work with your pdf. :)

[SeeMoreDigital]

Hi LordIntruder,

I think you may be correct regarding the default settings.

Even so, something can't be quite right if pressing on the default button alters the original installation settings.

Anyway, at least Manihi installation did not involve downloading and installing a new default settings registry file, like Kauehi!

Enjoy

[SeeMoreDigital]

OK. I've now encoded the same file using DigitAl56K settings. And I have to say there is an visible improvement in quality. So thanks for that.

I also performed the same encode again but this time disabled the Output Profile.

To confirm my results: -

Application Used: MpegMediator 1.5
Output Pixel Size: 720x576 Anamorphic
Codec Used: DivX Manihi
Bitrate Used: 627kbps
Number of Passes: 2

------------------------

1st Pass
Performance/Quality: Standard
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 50%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 300 frames
Selected Profile: Home Theatre
Bidirectional Encoding: Unchecked
1st Pass Encode Speed: 07Min 54Sec

2nd Pass
Performance/Quality: Slowest
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 50%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 300 frames
Selected Profile: Home Theatre
Bidirectional Encoding: Unchecked
2nd Pass Encode Speed: 58Min 26Sec

File Size: 29,109KB (28.42MB)

------------------------

1st Pass
Performance/Quality: Standard
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 50%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 300 frames
Selected Profile: None
Bidirectional Encoding: Unchecked
1st Pass Encode Speed: 07Min 54Sec

2nd Pass
Performance/Quality: Slowest
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 50%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 300 frames
Selected Profile: None
Bidirectional Encoding: Unchecked
2nd Pass Encode Speed: 58Min 28Sec

File Size: 29,109KB (28.42MB)

------------------------

Conclusion: -
Much better but unfortunately still not as good as WMV9 VCM. Even when viewing the output via the DivX Player 2.1

[SeeMoreDigital]

Is there any chance of enabling the old Mp4 output capability in future versions of the application?

Quite a few of us would like to create DivX quality Mpeg4 video streams with AAC / AAC+ audio!

[Darrius "Junto" Thompson]

Originally posted by SeeMoreDigital
Is there any chance of enabling the old Mp4 output capability in future versions of the application?

Quite a few of us would like to create DivX quality Mpeg4 video streams with AAC / AAC+ audio!

We can't do this from within the codec again. This caused quite a bit of problems both technical and product. Obviously one product issue is that the file won't be decodeable on and DivX Certified Hardware device so we don't want general users who don't understand this compatibility issue to create these files with the risk of thinking it would play on these devices. I haven't tried it in a while but I believe the MPEG4IP project had a command line app that was pretty fast at rewrapping a .avi to .mp4. Yepp just checked. Not sure how compliant it currently is but if they are keeping it up to date then it will might be more compliant than anything we might be able to leak out to you since we haven't updated ours in quite a few months.

Why the interest in the MP4 file format now?

http://mpeg4ip.sourceforge.net/faq/index.php#Divx

Darrius

[LordIntruder]

Hi SeeMoreDigital :)

I didn't even notice the "max keyframe interval" is set to 768 by default!!! I was a bit confused with the 80% to even notice this 768 figure! :(

Gosh, 768 on my PAL dvd 25fps = 30sec without any keyframe at worst.

I restore all the default config and now we retrieve our good old values. :) This is why you now get quality improvement over your first tests I suppose, you where with the default settings right?

[SeeMoreDigital]

Thanks for the quick reply Darrius,

My interest in MP4 is more to do with the container. I really like the DivX video format but I would like to experiment with AAC audio. Which is not possible with the AVI container.

In my opinion AAC audio and more importantly AAC+ audio will take off with a vengence in the coming months. We may even see (or should I say hear) 6Ch AAC/+ soon!

Quite a few people are generating AAC/+ test encodes at just 64kbps. And have reported favourable results. And since Mp3pro is unlikely to appear alongside encoding applications and players (even your own) then AAC and the Mp4 container seems the way forward.

As you are aware both RealMedia and WMV9 audio sounds alot better than Mp3 at 64kbps. And 6Ch audio is also possible (admittedly at 128kbps) with these codecs too.

DivX video is improving, your latest beta codecs prove that. But please don't forget about the audio. Not everybody wants AC3!

[DigitAl56K]

Remember that when quality testing, you must configure the decoder correctly. This might mean disabling the automatic post processing, setting full deblocking and deringing, and if you intend to scale the video (double size/fullscreen) you should have "overlay extended" and "yuv extended" enabled also so that your graphics card filters the output as it would for WM9 and that colourspace conversions are avoided.

[Gej]

It seems that there are some issue with the installer setting some default value, I suggest that you click on the "default settings" inside the codec before attempting an encoding, just to make sure that there is no issue (like the 768 frame keyframes interval that is, for sure, a little too much)

[bond]

in my tests

1) no mv file was written
2) i couldnt do a second pass - divx tells that there is something wrong with the source (the source image format isnt acceptable - error code -2)
i used (vob -> d2v -> mpegdecoder)

[SeeMoreDigital]

Hi DigitAl56K

Originally posted by DigitAl56K
Remember that when quality testing, you must configure the decoder correctly. This might mean disabling the automatic post processing, setting full deblocking and deringing, and if you intend to scale the video (double size/fullscreen) you should have "overlay extended" and "yuv extended" enabled also so that your graphics card filters the output as it would for WM9 and that colourspace conversions are avoided.

Well we have managed to view our DivX and WMV9 VCM encodes on a 42" plasma screen today via Sigma Xcard (using the Jove Player). And we also veiwed the same files on the same screen using our ATI cards VGA output.

On both occasions the WMV9 encode looked better at this bitrate.

I am happy to use any other suggested settings as they are found.
I am also happy to mail CD-R's of my tests to anybody who's genuienly interested.

[temporance]

@SeeMoreDigital,
In my tests, I find the quality of WMV9 VCM to be more variable than other codecs (can see thsi on a plot of PSNR). It also seems to start off with a higher quality which reduces as the clip goes on.... make sure you take a look at frames later in teh clip when comparing.

On your Mahini settings, what about Psychovisual?

[SeeMoreDigital]

Hi temporance,

In my first tests I've not used any psy. However I'm lucky that at work there are five of us (or more sometimes) that are able to give an opinion about an encode.

Many encodes are also generated on more than one PC. So there's no problems with say, codec clash on any given PC!

I'm going to try some more tests tomorrow with Manihi using the same keyframes interval as WMV9's 8000ms (8 seconds), which equals 200 frames.

Cheers

[DigitAl56K]

Ok, more information:

Gej tells me that the reason QPel is currently running so much faster is that the new RD algorithm is not enabled for QPel.

This means that slow/slowest modes (which feature the RD algorithm) are only available for half-pel, and that you will have to put up with slow encoding for now in order to test them.

If you have been testing with QPel and experienced a lower quality than WM9, this may be why.

[SeeMoreDigital]

OK. I've generated some more DivX 'Manihi' files today using settings that are as close to M$'s WMV9 VCM encoder as I could.

My results are as follows: -



Source 'Video Only' File Information

Source 'Video Only' File Information
Selected Movie: StarWars 2 PAL
Selected Chapter: 41
Clip RunTime: 06m 17s (377.24 secs)
Total Frames: 9431
Pixel Size: 720x576
Video Type: Mpeg2 (M2V)
Field Type: Progressive
Output AR: 16:9 (1.77:1)
Image AR: 2.35:1
File Size: 243,030KB (237.33MB)

-------------------------------------------------------
-------------------------------------------------------

DivX Encode Information

Application Used: MpegMediator 1.5
Output Pixel Size: 720x576 Anamorphic
Codec Used: DivX Manihi
Bitrate Used: 627kbps
Number of Passes: 2 VBR


First Pass
Performance/Quality: Standard
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 50%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 200 frames
Selected Profile: None
Bidirectional Encoding: Unchecked
1st Pass Encode Speed: 07Min 54Sec


Second Pass
Performance/Quality: Slowest
Feedback Window: Disabled
Show Picture: N/A
Psychovisual Enhanc: Disabled
Enable Crop: Unchecked
Enable Resize: Unchecked
Pre Processing Source: Unchecked
Scene change threshold: 50%
Source Interlace: Encode as Progressive
Max Keyframe Interval: 200 frames
Selected Profile: None
Bidirectional Encoding: Unchecked
2nd Pass Encode Speed: 59Min 10Sec

File Size: 29,107KB (28.42MB)

-------------------------------------------------------
-------------------------------------------------------

WMV9 VCM Encode Information

Application Used: MpegMediator 1.5
Output Pixel Size: 720x576 Anamorphic
Codec Used: WMV9 VCM 'Latest Release'
Bitrate Used: 615kbps
Number of Passes: 2 VBR


First Pass
Under 'Compression' Tab: -
Decoder Complexity: Main
Performance: Full 5 'Better Quality'
Max Keyframe Interval: 8000ms (200 frames)
Log File: First pass - General info added

Under 'Pre-processing' Tab: -
Source Mode: Encode as Progressive
Frame rate downsamp: 1:1
Resize to: Unchecked
Enable cropping: Unchecked

Under 'Save/Load' Tab: -
All Presets: Not Used
1st Pass Encode Speed: 09Min 38Sec


Second Pass
Under 'Compression' Tab: -
Decoder Complexity: Main
Performance: Full 5 'Better Quality'
Max Keyframe Interval: 8000ms (200 frames)
Log File: Second pass - General info added

Under 'Pre-processing' Tab: -
Source Mode: Encode as Progressive
Frame rate downsamp: 1:1
Resize to: Unchecked
Enable cropping: Unchecked

Under 'Save/Load' Tab: -
All Presets: Not Used
2nd Pass Encode Speed: 74Min 54Sec

File Size: 29,096KB (28.42MB)

-------------------------------------------------------
-------------------------------------------------------

Conclusion: -
Sorry, but at this bitrate, in my opinion (and the opinion of my co-workers) the encoded image does not look as good as WMV9 VCM. That's not to say it's poor, because it's not poor by any stretch of the imagination. Indeed it far exceeds the current version of DivX Pro 5.0.5 and produces images that look slightly better and encode faster than Kauehi. It even encodes faster than WMV9 VCM!

And as I mentioned before I really like the new GUI. I'm very pleased with the results and so is my Xcard. And who knows, when Sigma release their new drivers (which will be available very soon by all accounts) the images on TV screen may look even better!

Now, I realize ofcourse that not everybody likes to encode at low bitrates. And once you get above 950-1000kbps DivX Manihi really shines. Indeed at 1500-2000kbps a 2pass VBR encode looks better than some first generation DVD's and even most of SkyDigital's output in the UK today!

Great work you guys and thanks!

[temporance]

@SeeMoreDigital, a couple of questions:

A closed question... what postprocessing are you using for your comparison?



And a leading question... if I was an untrained viewer, which video would I like the best?

[Untrained viewers aren't sensitized to blocks like us "pros" and prefer the sharpest looking video, even if it has more blocking because [i]blocks look sharp -- this explains your opinions of Sky Digital!]

Edit: another question: what's the CPU load of playback for DivX Manihilate and WMV9?

[Fr4nz]

Why you don't use bidirectional encoding with Manihi?? IMHO It could enhance the quality...

[SeeMoreDigital]

Hi temporance

A closed question... what postprocessing are you using for your comparison?

I (or should I say we) have tried the following post-processing settings: -

With DivX
Deblocking: Max / Deringing: Max. And we've disabled Post Processing too.

With WMV9
We've tried post-processing at default. And we've disabled it in the registry

I understand what you mean about 'untrained' viewers. But that's not so much the case with me as I've been in this field of work for over twenty years now. Sure, I don't confess to know everything, (I'm more a hardware man myself) that's why we are all here on the forum.

That said, I agree that it's a common mistake for people judge their encodes by sitting too close to their monitors. But as I said before that's why we view them on a 42" plasma screen, in the correct light and use a variety of different PC - plasma output connections.

Now, when it comes to the SkyDigital quote. I think it's fair to say that the quality of the output (BBC1, BBC2, ITV1, ITV2, Ch4 and Five) has dropped. I've noticed this more and more over recent months and when I switch over to my terrestrial digtal box and watch the same output the image appears to be better.

Shame really, because it used to be the other way around. My only conclusion is that the drop in SkyDigital's quality is due to the terrestrial broadcasters trying to squeeze more regional channel variations out of the same transponders!

Oh, and as for the CPU load. For some reason WMV9 VCM uses less. Which is strange because when the same file is encoded as an .wmv stream it uses more. More even than DivX.

MingCL at M$ is aware of this, as I posed this question to him on a previous thread.

I'll mail you some of our tests if you like!

[temporance]

By "untrained viewers" I meant people that just like to watch TV and movies, not compress them or talk about video quality. These are the lowest common denominator that make up 99% of Sky's customers. They'd prefer blocky to soft.

[SeeMoreDigital]

Ha Ha,

I'd have to agree with you there.

But what can SkyDigital or any other digital customers do? Back in 1998 they were given a 'quality' product to start with. Todays decoders are better than ever and yet the quality of the broadcasted image is poorer than ever.

It's a conundrum, a double edged sword, six of one and half a dozen of the other. It's a pain in the ar*e, is what it is!

[Sumster]

Just finished encoding "How to Lose a Guy in 10 Days". It was a 640x480 (4:3 fullscreen) encode using lancos resize. 2-CD rip so the bitrate was almost 1500. B-frames on, GMC off, PVE off. Turn on WRITE MV on first pass and then READ/UPDATE MV on second pass. Used the "SLOW" setting for both.

1) Averaged about 2.0 fps on my Athlon 1800+ with 512 mb.... took about 23 hours per pass (total of 45 hours for both passes).

2) No speed increase on the second pass :(. No MV file was written so I am not sure if my settings should have been different. I assumed the second pass would be much quicker by re-using the motion vectors.

3)Quality was phenomenal... the best quality rip ever. PSNR calculations to follow.

4) Love the dialog box (I had picture and graph DISABLED during the encode). What does "SAD" mean? What are the grey boxes in "Compensated Mode"?

Take care..

Sumster

[Sagittaire]

My personal test: Visual and PSNR test with RV9 EHQ, XviD, DivX, WMV9

http://forum.doom9.org/showthread.php?s=&threadid=57687&perpage=20&pagenumber=1

Manihi standard mode is better than XviD Koepi 24/06/03 ...

SAD is an fonction of Motion Estimation.

[LordIntruder]

temporance:

"untrained viewers"....They'd prefer blocky to soft

I don't see why untrained users would love sharp videos and so call pro would love blurry ones? You prefer a blurry video that means blurry details, etc... rather than a sharp image? Well I never use post-processing because I don't want a modified video. I do not use any post-pro to blur nor sharper it. Unless the video is bad, macroblocked, etc... so I may used post-pro to blur it a bit otherwise not.

And I don't understand neither why people compare the different encodes with the post-processing mode on. To my point of view we have to compare encodes without these tools to see the rough differences between codecs and after put post-processing on and compare again and in the final report say when post-pro was and wasn't used.

[Fr4nz]

@ SeeMoreDigital: please could you try to make some tests with B-Frames activated and then tell us the results?

Oh and tell us also if the encoding speed is lower...!

[V-tec]

Menihi produces a very good quality encodings, but is good for test only, (20+20+20+...+n) min to encode 4 min of video is not good...
I hope that in final version it's run faster

[Sagittaire]

Manihi standard mode is a very faster mode (same speed of DivX 5.05) ...