Manual way to make a comp. test with DivX (can be used with other codecs too)

[jonny]

I suggest a manual method to make compressibility test with DivX:
(This method can be applyed to every codec)


1 - Check "Compressibility Check" in GKnot and save your avs file.

2 - Open the avs file in VDub, select 1-pass quality based and set the quantizer to 2 (set the other options as you want).

3 - Save to an avi file.

4 - Open the avi file, go to File > File Information.

5 - Take a look at the row "Min/avg/max/total delta frame size", take a note of the "avg" value.

6 - Multiply "avg" by the total number of frames in your movie, so you have the "predicted size" at quality based 100%.

7 - Take a note of the "Video Size in KB" in GKnot, you must multiply this value by 1024 to obtain the size in bytes.

8 - 100 * "Video Size in bytes" / "predicted size" is an approximation of the comp.test value.


Another use of the method: "Size prediction":
If you use a different quant at step 2 and you stop at step 6, you'll obtain the predicted size for the quality based encode at a given quant.



There is an Excel spreadsheet that can be very useful for making the calculations:
http://peque.homeftp.org/~maurus/comptest.xls
It's in Spanish, anyway it's realy easy to use.
Many thanks goes to maurus! :)
killingspree have translated the spreadsheet in english (many thanks to you too!):
link1: http://members.tripod.de/stefanstrobl/comptest.xls
link2: http://jonny.leffe.dnsalias.com/doom9/comptest.xls


This method is not 100% accurate with b-frames enabled.
With b-frames the predicted size will be oversized and the resulting comp.test value will be lower than the real value (i think something like 10%-15%).

I hope you don't get confused with this manual method, note that DivX5Enc method and GKnot method are 100% accurate with b-frames enabled

Thanks goes to Elia for this nice manual method (that work with every codec)! ;)

[drakoh]

thx buddy, it is good to know !

[hakko504]

*Sticky*

[hakko504]

@Jonny

Could you insert the 'save avi' step between 2 & 3 . I think it will be very har getting any information out of a non existent avi ;)

[jonny]

:D
Done.

[N_F]

I've been thinking a lot about it and I can't quite figure out why this method wouldn't give correct results for B-frames. I realise that the keyframes will make it less accurate (though that you can take into your calculations), but why B-frames?

The only thing I can think of is if VDub don't quite take B-frames into calculation when it reports avg. delta frame size. Would that be it?

[jonny]

N_F, take a look at my post here:

http://forum.doom9.org/showthread.php?s=&threadid=29357

[N_F]

I just read it and I think I understand the basics of it. A small suggestion: You might consider adding so the predicted size will be oversized and the resulting comp.test value will be lower than the real value (i think something like 10%-15%), so people can take it into their calculation (assuming the number 10-15% isn't 99% guessing from your side).

[jonny]

Nice suggestion, done.

[hakko504]

@binky
Interesting question, but you are way off topic in this sticky. Now, there is no explicit rule against threads drifting off in directions never intended by the original poster, but it is my strong opinion that we should try to keep STICKY threads as clean as possible. Please move your question to a separate thread, and I'll be glad to give you a comment.

Binky has now moved his post in this thread to a new thread, but I will let my answer remain as a reminder that when you post in a sticky, please keep the focus as much as possible on the original subject. Thank you.

[Redin]

Please i do not understand very well what are you mean about your last information.

Did it possibele use automatic compressibility check with new divx 5.0.3 and GordianKnot?

what is a Quantizer and the pass 1 your wrote what do you mean, please?
If you are referred about divx5.0.2 is ok. But in divx5.0.3 quantizer do not exist.

[jonny]

You must disable profiles, so the 1-pass quality based mode will be showed in the list, then, it's all very similar to DivX 5.0.2.

Atm it's GKnot doesn't support 5.0.3 :(

[nan1nan1]

hi
in step 7
i dunno where i can find "Video Size in KB in GKnot"
is it just the filesize without audio?and i can simply get it in Windows Explorer?
one more silly question :p
what kind of result did u get?
i got something around 16 when i use the filesize as the "Video Size in KB in GKnot"

Regards.:)

[jonny]

You can find "Video Size" in "Bitrate" tab of GKnot.

16 is possible, but you should give more detailed informations to make me understand if you are doing something wrong.

Cheers
jonny

[nan1nan1]

i think the bitrate tab is only useful after reading 1pass .stats file.
follow your tips,i encoded using the 1 pass quliaty 100% mode...no .stats file generated.that's y i dunno where to find the video size in GKnot.

and i just did an ordinary encoding.only check the "Compressibility Check",set everything else default..

Regards.

[jonny]

I think you have some confusion about this method.
I've attached an image about the video size in GKnot.
The video size is determinated before you make a comp.test, it depends by the length of the movie, audio and total target size you set (1cd, 2cd ecc...)

[Techguy]

I can't seem to get this to work.
when I pull up the file info after the compressibility check, It says no delta frames
http://www.prism.gatech.edu/~gtg380h/images/vdub.jpg
I did turn off profiles in the 5.03 options and set the pro features that were reccomended on the Divx5 guide here. (for 5.02 i know).

any ideas?
This is a fresh format PC (just reformatted) and I just installed Gknot and saw the new DivX codec.

Thanks

[jonny]

when I pull up the file info after the compressibility check


4 - Open the avi file, go to File > File Information.

Are you doing this?

[Techguy]

Originally posted by jonny
4 - Open the avi file, go to File > File Information.

Are you doing this?
Ah, I did the file info on the avs, not the AVI.
thanks!

[killingspree]

i've just done a quick translation of the spanish Excel spread sheed... should be attached below... well has to be validated first :-)


regards
steVe

[jonny]

Thank you, i've updated the sticky.

[theReal]

As there is now an option for quant 1 in Divx 5.1 - do I still choose quant 2 for the (manual) compress test?

(I know there has been a hidden way to use quant 1 in previous versions of Divx5, but I was wondering if quant 2 now still means "qaunt 2" in Divx 5.1 or if the old quant 2 is now called "quant 1")

[jonny]

Hi theReal ^^


but I was wondering if quant 2 now still means "qaunt 2" in Divx 5.1 or if the old quant 2 is now called "quant 1"

A quantizer is always a quantizer :) (at least if DivX doesn't decide to use the word with a different meaning from the mpeg4 standard).

As there is now an option for quant 1 in Divx 5.1 - do I still choose quant 2 for the (manual) compress test?

You can really do what you want. Changing the quantizer only change the "scale" of values you obtain from the compressibility test.

I make an example:
If with quant=2, your target compressibility value is 60%
using quant=1, your target could be for example 40%.
You could even use quant=3 and decide that your target is a comp.test=100%.

Comp.test is a simple value: target_size/predicted_size_with_a_fixed_quant

So choose a quantizer and make some tests to see at what comp.test value, the movie looks good to you.

Anyway, the simple answer is: using quant=2 you'll probably obtain values that are "compatible" with the previous versions (*but* warning because this is not completely true with 5.1 - testing to see what is the best for you is, again, the only choice).

[theReal]

Ok, thanks!