New Highly Accurate Prediction Method Reported by Jorel! - Page 3

Krassi · #41 07-15-2003, 02:01 AM

Quote:

Originally Posted by jorel

someone got good news results or
bad news results with this prediction way

please,...all hints, errors, goods...all results are welcome.
the target is find the best way!

Hi Jorel,
my experiences:
I had better results with a sampler length of 24 on (german) PAL sources than 48 and 72.
48 for me is better than 72 but only sometimes worse than 24.
With the "normal" PAL-SL of 25, 50, 75 i don't get accurate prediction. With SL=24 or 48 my final size is within 1% all the time

.
I've checked this with normal and fast prediction and it doesn't make any difference.
My last encoding was only 600 KB off target

.

Currently i'm testing Variable Bitrate (2-pass). The encoding was good, and wanted average bitrate was exactly encoded average bitrate. But encoding was very slow (11 hours on a 2000+)...

kwag · #42 07-15-2003, 02:43 AM

Quote:

Originally Posted by Krassi

Currently i'm testing Variable Bitrate (2-pass). The encoding was good, and wanted average bitrate was exactly encoded average bitrate. But encoding was very slow (11 hours on a 2000+)...

Why do you do that Krassi

Are you aware that you'll never get the same quality as with CQ

At least with TMPEG

-kwag

Krassi · #43 07-15-2003, 02:50 AM

Quote:

Originally Posted by kwag

Why do you do that Krassi

Are you aware that you'll never get the same quality as with CQ

At least with TMPEG

Yepp i know

. I'm just doing some experiments with the mainconcept encoder/TMPGEnc and trying to tweak and compare settings.
The quality was a litte worse on my PC (CQ vs VB) but i didn't see any difference on my TV. I think i need glasses

.

jorel · #44 07-15-2003, 10:33 AM

great results Krassi,
thanks for post!

anyone more?
c'mon boys here is the Kwag's kvcd forum,
we "never say die"

gonzopdx · #45 07-15-2003, 05:53 PM

So I'm assuming that using this new prediction method (using 3 * frame rate), if I'm using KVCD Predictor to get my target file size, I want to use triple the sample points (or just triple the sample size)? Right?

Avalon · #46 07-16-2003, 06:13 AM

So which are the best settings for PAL DVD Sources?

Are they these in ToK 0.0.5.3?:

Main:

Min. Bitrate 300
Max. Bitrate 2500
Initial CQ 80,00

MPEG 1 multipex to VCD

Prediction & Spacing:

Factor for Predicition: 1,00
Length of Sample : 75
Number of Samples/minute: 1

Faster Prediction: unchecked

1st Group of Passes:

Precision %: 0,50
Speed Up by: 10

2nd Group of Passes:

Precision%: 0,50
Speed Up by: 1

Fixed: Number of CDs 80min

Audio & More:

External Audio: 'encoded Audiofile by BeSweet'

TMpegEnc Parameters: 1, 12,2,1,25

Have anyone other, better settings to get the closed precision for an 80 Min CD?

Avalon · #47 07-16-2003, 07:10 AM

Here the ToK-Log-file:

Quote:

================================================== ===========
ToK Log: D:\dcd\blade\blade.avs
================================================== ===========

Resolution (fps):352x288 (25,000 fps)
Total Frames: 138824
Total Time : 01:32:33
-------------------------------------------------------------

Audio Size: 111.060.114
Required Video Size: 640.232.887

Factor: 62,500
Desired Sample Size: 10.243.726

-------------------------------------------------------------
Prediction: 1st Group of Passes
-------------------------------------------------------------

Next CQ: 81,910. Sample Size: 9.400.030
Next CQ: 89,262. Sample Size: 10.872.520
Next CQ: 85,854. Sample Size: 10.823.460
Next CQ: 84,065. Sample Size: 10.749.590
Next CQ: 83,122. Sample Size: 10.199.400

Exit Condition: 0,500 % reached ! yahoo !
Tries : 5

-------------------------------------------------------------
Prediction: 2nd Group of Passes
-------------------------------------------------------------

Next CQ: 83,122. Sample Size: 11.491.243
Next CQ: 74,098. Sample Size: 7.063.375
Next CQ: 79,595. Sample Size: 8.138.876
Next CQ: 81,584. Sample Size: 10.357.445
Next CQ: 81,036. Sample Size: 9.828.133
Next CQ: 81,388. Sample Size: 10.211.454

Exit Condition: 0,500 % reached ! yahoo !
Tries : 6

Final CQ: 81,388
Total Time For Predicition: 00:16:33

-------------------------------------------------------------
Encoding D:\dcd\blade\blade.avs
-------------------------------------------------------------

Encoding... CQ : 81,388
Final Encoded Size: 618.799.157
Muxing ...

Total Time (all operations): 02:23:36

Finished

Audio Size: 111.060.114
Required Video Size: 708.139.886

Factor: 20,000
Desired Sample Size: 35.406.994

-------------------------------------------------------------
Prediction: 1st Group of Passes
-------------------------------------------------------------

Next CQ: 80,000. Sample Size: 24.106.470
NextCQ: 117,502 >=90, so check is needed
Next CQ: 90,000
Sample Size: 34.441.840

NextCQ: 92,522 >=90 once again, a more detailed check is needed
Next CQ: 90,000
Sample Size: 39.392.658
-------------------------------------------------------------
Prediction: Special, CQ more or less 90
-------------------------------------------------------------

Next CQ: 80,894. Sample Size: 31.365.482
Next CQ: 85,463. Sample Size: 39.212.085
Next CQ: 83,213. Sample Size: 37.251.953
Next CQ: 82,270. Sample Size: 35.179.958
Next CQ: 82,557. Sample Size: 35.834.956
Next CQ: 82,391. Sample Size: 35.431.373

Exit Condition: 0,500 % reached ! yahoo !
Tries : 8

Final CQ: 82,391
Total Time For Predicition: 00:50:57

Total Time (all operations): 00:50:57

Finished

I didn´t reencode this movie, because i find CQ 81,388 is fine...

Here the used AVS-Script:

Quote:

#================================================= =====#
# -= AviSynth script by MovieStacker v2.0.0 (beta3) =- #
#================================================= =====#

LoadPlugin("C:\Programme\MovieStacker\Filters\MPEG 2Dec.dll")
LoadPlugin("C:\Programme\MovieStacker\Filters\STMe dianFilter.dll")
LoadPlugin("C:\Programme\MovieStacker\Filters\Grip Fit_preview.dll")
LoadPlugin("C:\Programme\MovieStacker\Filters\Bloc kBuster.dll")
LoadPlugin("C:\Programme\MovieStacker\Filters\UnFi lter.dll")
LoadPlugin("C:\Programme\MovieStacker\Filters\Lega lClip.dll")

Mpeg2Source("D:\dcd\blade\bl.d2v")
LegalClip()
GripCrop(352, 28

GripSize(resizer="BicubicResize")
STMedianFilter(8, 32, 0, 0, 8, 32)
UnFilter(50, 50)
TemporalSmoother(1, 1)
MergeChroma(blur(1.5))
MergeLuma(blur(0.2))
BlockBuster(method="noise", detail_max=7, seed=1)
GripBorders()
LegalClip()

The movie was 'The Princess Blade'.

Now i will try it with the movie Lost in Space...

Avalon · #48 07-16-2003, 01:56 PM

With this way i got the final size 100,1 %!!!!!

Code:

ToK 0.0.5.3?:

Main:

Min. Bitrate 300
Max. Bitrate 2500
Initial CQ 80,00

MPEG 1 multipex to VCD

Prediction & Spacing:

Factor for Predicition: 1,00
Length of Sample : 75
Number of Samples/minute: 1

Faster Prediction: unchecked

1st Group of Passes:

Precision %: 0,50
Speed Up by: 10

2nd Group of Passes:

Precision%: 0,50
Speed Up by: 1

Fixed: Number of CDs 80min

Audio & More:

External Audio: 'encoded Audiofile by BeSweet'

TMpegEnc Parameters: 1, 12,2,1,25

Here the ToKLog:

Quote:

================================================== ===========
ToK Log: D:\dcd\lost\lost.avs
================================================== ===========

Resolution (fps):352x288 (25,000 fps)
Total Frames: 187263
Total Time : 02:04:51
-------------------------------------------------------------

Audio Size: 149.812.767
Required Video Size: 669.387.233

Factor: 20,000
Desired Sample Size: 33.469.362

-------------------------------------------------------------
New Faster Prediction
-------------------------------------------------------------

Full Sample
Next CQ: 80,000. Sample Size: 39.637.921
Small Sample
Next CQ: 80,000. Sample Size: 3.223.735
Predicting...
Next CQ: 67,550. Sample Size: 26.742.587
Next CQ: 73,910. Sample Size: 32.251.678
Next CQ: 75,934. Sample Size: 32.534.798
Next CQ: 77,218. Sample Size: 32.675.460
Next CQ: 78,072. Sample Size: 32.772.399
Next CQ: 78,652. Sample Size: 32.849.136
Next CQ: 79,051. Sample Size: 34.433.517
Next CQ: 78,830. Sample Size: 33.237.888
Next CQ: 78,906. Sample Size: 33.722.754
Next CQ: 78,867. Sample Size: 33.442.610

Exit Condition: 0,500 % reached ! yahoo !
Tries : 11

Final CQ: 78,867
Total Time For Predicition: 00:19:50

-------------------------------------------------------------
Encoding D:\dcd\lost\lost.avs
-------------------------------------------------------------

Encoding... CQ : 78,867
Final Encoded Size: 669.970.750
Muxing ...

Total Time (all operations): 02:59:16

Finished

The final size was exactly: 801,00 MB!

ToK says 669.970.750 and he wanted 669.387.233. Pretty close!

jorel · #49 07-17-2003, 09:46 PM

thanks for extreme details Avalon.

i am(we all) happy with your results too

Bilal · #50 07-23-2003, 07:41 PM

i followed avalon's details and i left TOK running overnight thinking i'll have the movie ready by morning and in the morning it was still predicting hehe maybe because i have an 800mhz processor.

jorel · #51 07-23-2003, 07:51 PM

hey Bilal

try the CQMatic 1.0b5 that give
fast and precise predictions.

details and download:

http://www.kvcd.net/forum/viewforum.php?f=64

rhino · #52 07-24-2003, 04:13 AM

@Jorel:

I tired you method in Tok with 4 Pal DVD's and so far they have been pretty close to perfect. After muxing they range from 765Mb to 785Mb. These where at the 352x388 res. so i'll move onto the next res and see how that works out.

A good improvemet!

Cheers,

Zyphon · #53 10-13-2003, 06:47 PM

I know im a noob but i thought id ask my silly question any way lol.

Will this method work on AVI encoding or is it just for DVD source only?

Thanks in advance.

Dialhot · #54 10-14-2003, 06:17 AM

This is a really old thread. Things change a lot here. You can use last CQMatic for accurate prediction even with avi (but using a static script, not the MA one).

Zyphon · #55 10-14-2003, 09:25 AM

Thanx 4 ur reply Dialhot i understand now.

Much appreciated.