I guess we skin the pig ehhh :mrgreen:

Edit: you all did :!: :mrgreen:

[quote="kwag"]It's Miller Time :!: woot. Starting to run my tests and I can't wait to see the results. :D :D :D

Here you go guys :)
Here's the small sampler of K19 (~1MB), so that's how the whole 138 minutes looks like on one CD-R :mrgreen:

http://www.kvcd.net/k19-fast-predict...-cq-63.209.mpg

Enjoy :)
-kwag

So do we continue using tok 0.0.5.2 or can we go back to 0.0.5.3 and what settings should we use?



David

Hi David,

I havent tried that yet :!:
I'll give that a shot too, and see if the CQ matches a very close CQ as 0.0.5.2.
If it does, then we can use 0.0.5.3 :)
I'll run both predictions. The full, and Tenra's "Fast" prediction.
Only if both match, I would then use 0.0.5.3.

-kwag

Here's the result with ToK 0.0.5.3, which uses "SelectRangeEvery()" instead of "Sampler()" for prediction.
This was on my "K-Pax" movie.

CQ found using "Tenra's" Fast prediction: 70.138
CQ found using normal (long) prediction: 69.940

The CQ found by ToK 0.0.5.2 was 65.51 using Tenra's "Fast" mode, and the full prediction was almost the same number. And that is the correct CQ, because I left the movie encoding last night, and the result this morning was +1.73% accurate :)

This proves that the "SelectRangeEvery()" is a problem for prediction, but SansGrip's method of selecting frame range is excelent :)
So I would stick to 0.0.5.2. Other than the file prediction method used, there's no difference from 0.0.5.2 and 0.0.5.3

-kwag

Hi Guys, last night i encodes (A guy thing) 1:41 min 70.61% CQ on one CD output video was 710MB, audio was 83MB for a total file size of 793MB.
This was using tok 5.2 n tenras fast prediction, so i guess this turkey is really cooked now. :D

Hi ovg64,

What was your % accuracy on that encode :?:

-kwag

and please Osvaldo,
write all details,..if is mpge1 or 2,the "resize",time of the source,
everything cos i loose the high prediction again,
then i need of all informations.
my CQ back to 56 "house" again.

Hey Guys, here you go; :D

Mpeg 1
712 mb video wanted 710 obtain
83 mb audio at 112 kbps
1, 24, 2, 1, 24 gop

Fast Prediction *
Factor 1.00
Sample Auto *
# sample/min. 1

1 Group Pass
presicion % 0.50
speed up by 10
10:00 % presicion

80min. CD -1

ovg64: That's great that you got such an accurate prediction! I've been wondering lately at what point should we predict using 2 samples/min instead of the default 1? I've pretty much been raising it to 2 whenever the movie is shorter than 2:00:00 long, but I see this one is 1:40:00. Any opinions?

I dont know but it sims like sample() is more accurate than selectevery()
use in tok 5.3 at least in Avs 2.52, cause yesterday i use 3 setting mine, kwags n Tenras and they all came to around the same 70% CQ for this same movie. :)

really strange Osvaldo.
my source:

Resolution (fps):480x480 (29,970 fps)
Total Frames: 159173
Total Time : 01:28:31

Audio Size: 84.967.000
Required Video Size: 734.224.000

:arrow: and my best CQ56,255!

why i with: less resize, less time movie got less CQ :?:
:? :? :? :?

i need a BIG help :!:
:(

What are you ecoding :?: National Security :?: :D 1:28:31 time
post your script i think maybe one of your filter might be doing
this are u using DGbob :?: you are using 29fps that might be another
reason. :!:

no Osvaldo,no DGBob no more i cahnge to fielddeinterlace!

see my script:

MaxTreshold=1.50
nf=0 # Current frame.

Mpeg2Source("D:\La Luna\Segment2\DVD2AVI_PROJECT_FILE.d2v")

FieldDeinterlace()
undot()
Limiter()
asharp(2,4)
GripCrop(480,480,overscan=1,source_anamorphic=fals e)
GripSize(resizer="BicubicResize")
STMedianFilter(8,32,0,0)
MergeChroma(blur(MaxTreshold))
MergeLuma(blur(0.2))
ScriptClip("nf=YDifferenceToNext()"+chr(13)+"nf>2. 5?\
unfilter(-(fmin(round((nf/0.5)),100)),-(fmin(round((nf/0.5)),100))):\
TemporalSoften(2,7,7,3,2)")

GripBorders()
Limiter()

function fmin(float f1,float f2){return(f1<f2)?f1:f2}

thanks in adbvance!
:wink:

ps:
Jell is helping me too in the "best scripts" thread...
of course,something is wrong in my system and
i don't know what is ?!?!?
:?

@jorel

Look at the Optimal Scripts forum... :wink:

Hey Jo your script looks ok to me, i dont know maybe tok doesn't pred. well with 29 fps. Maybe you want to experiment a bit i would maybe put undot after the resize n take out the deinterlance filter just to see what kind of CQ you get or maybe try another deinterlance like tomsmocomp
for 2.52 and experiment a bit you know im not a brain like you but i some how solve my problems :wink:

I can confirm this for PAL-Sources. Prediction is very accurate for PAL using ToK 0.0.5.2, a Sampler Length of 24 and the newest adaptive script, even with Tenras new prediction. The prediction was 0,3 % accurate on a middle action movie.
Thanks to Kwag, Jellygoose and all others for helping figure this out.
Maybe we should contact Pegasys-Inc. to determine if theres a bug in PAL-GOP encoding :roll: .

Does this mean that we can continue using v0.0.5.3 and just substitute
SansGrip`s "Sampler"??? (sorry if this has been answered already)

Or do we need v0.0.5.2,<---(???) If so...where can we find that version??


*******************************
The Devil`s always.....in the Details!

@ Jell
thanks, i read all there and waiting your helper answers. :)

@ vico1
good question,i want to know the same.
:wink:

@ Osvaldo :)
tomsmocomp is worse,decrease the CQ.
i change deinterlance to just before and just after resize and change just a little...insignificant.
fielddeinterlace is in the better place in the script,
(thanks to Boulder and Phil for big explanations about deinterlace and ivtc)

you wrote:
"...i'm not a brain like you..." 8O
you are in better perspective,you have a good brain working,
mine is lost. :roll: