Quote:
Quote:
-kwag |
I think we could just up our prediction factor by a smidge if doing fast prediction, no? Like say 1.01 or 1.02
|
I keep coming up 5 or 6% short even though as you guys said the CQ values are close. I'm going to keep trying, but I think 48 frame samples at 2 per minute will really slow down prediction time, but we'll see. I haven't tried yet.
|
Quote:
So you could end with -30MB or with +30MB, depending if the sampler was -0.5% or +0.5% from wanted file size. I rather do a full prediction with a factor of 1.0 just to be sure. I'm currently starting to encode a complete different movie I haven't tried with the latest settings. It's "Under Siege 2". We'll see tomorrow how it comes out :) -kwag |
@ Kwag
I'm encoding "Foreigner" Pal version and have 2 problems 1 . Encoding with last script is taking 8hr 20 min and with script before last changes 3hr 35 min CQ = 61 in both cases and res=704x576 with last script quality is slightly better (less mosquitos around subs ) but encoding time ~3 times slower ???? 8O 8O 8O What could be the problem ??? 2 . Prediction with Tok 0.0.5.2 or 0.0.5.3 with fast mode always results files 40-50mb smaller then whanted . I tryed to play with param's but cant get constant results with different movies . bman |
Hi bman,
For me ToK never worked correctly, so I tried another procedure which gave me correct results for PAL, but it is not wide tested. I will give you a link to the post http://www.kvcd.net/forum/viewtopic....=asc&start=128 Try it, if you want. |
Quote:
I'll try that as soon as I'll arrive home :D :D :D bman |
Quote:
have you tried ToK 0.0.5.2 with a Sampler length of 24 or 48 :?: My predictions are within 1% since i use these settings (doesn't matter if i use faster prediction or the older one). You could also try to go for a higher value at samples/minute. |
Quote:
I've try 48 :D Thank's Krassy bman |
Here's my result on "Under Siege 2":
Code:
============================================================= -kwag |
Thank's Kwag !!!
I'm encoding with 704x576 and it takes ~8.5hr to encode . U have encoded 528x480 and it took ~5.25hr to encode . So 704x576 / 528x480 *5.25 = 8.4 hr It's about the time that took me to encode "The Foreigner" with CQ 61 . I'm happy that my system is tuned up and quality is great too but 8.5hr encoding is too much time for me , on job that could be done in 3.5hr with previous script and hardly noticable difference :? :? :? Correct me if I'm wrong :? :? :wink: bman |
Quote:
-kwag |
Quote:
We just missunderstand one each other :D I'm talking about MA script before the last changes u made (with conditionalfilter and without ) With NEWEST script If u are encoding 528x480 it could be a big difference but not with 704x576 . Even on 34" TV it's hard to see difference cos TV is smoothing most of mosquitos . That's just my opinion , That's what I see :wink: :wink: bman |
Script Screwed ?
Somethins wrong with jul 1st script 2.52
(Keep gettin white writing on the screen along with my movie that i am encoding) |
Re: Script Screwed ?
Quote:
Sykes,download temporalcleaner here: http://www.avisynth.org/~warpenterprises/ i think that you don't have this filter for use in the new script! :wink: editing: no need the filter no more, temporalcleaner was changed for temporalSoften. copy the script again! :wink: |
Re: Script Screwed ?
Quote:
Change this line: ScriptClip("nf = round(YDifferenceToNext()+1)"+chr(13)+ "nf > 2 ? \ Dialhot's line read: ScriptClip("nf = round(YDifferenceToNext())"+chr(13)+ "nf > 2 ? \ and the value would go to zero, which causes "division by zero". With the above fix, now it never goes below 1 ;) -kwag |
Looking at the changes by Dialhot, I overlooked something :!:
The values are integer now, but that throws OFF completely the boundary between the switching of blurring and temporal filter. Before, there was a floating point precision. And now it's sharp values that jump from 1 to 2, and that's NO good :!: I'm revising the script again :!: :!: Edit: Ignore this post. I need sleep. The script is correct :oops: -kwag |
:wink:
|
hi Kwag.. :)
Quote:
good in PDVD v3.0 !! Please keep it up :lol: Hay, we're gonna have a LONG weekend this week. Our company is letting us go Thur and Friday !! And, I've got things cooken for this week/end.. See you then.. ..TGIW, almost here hehe.. Be well all. -vhelp |
I'm seein blocky areas in my encode anyone suggest what I can change or add to the optimal script datetd 1 july?
|
Quote:
i see the same but less "breathing" in each second! just a "little few" beating changing every second! little blocky areas too like andybno1 posted. only i see it? nobody see? :?: |
Quote:
But yo can try round(2/nf) insteed of round(3/nf) in the third parameter of temporalsoften. |
sorry?
why? post how you use temporalcleaner, i want to test it too! :wink: |
I think this is what Dialhot uses Jorel. :idea:
Quote:
|
thank you Osvaldo!
:wink: what's your results testing with temporalcleaner? |
Phil and Osvaldo:
with temporalcleaner the colors are draw...seems better, the size with my sample encrease a little but straight "breathing" yet...like pulsing,beating :!: :( |
hi Jo, i haven't really try TemporalCleaner in the MAS yet, but that pulsing that you are talking about sounds maybe like your memory defrager setting is too high, just a guess dude. :wink:
|
yes,Osvaldo..maybe this is the problem of "beating".
send me recomendations...what i do? :? |
Quote:
|
8O
i don't know how to do it. please Osvaldo pm to me, i am "off topic". :) |
You can PM that to me too, I think it's maybe off-topic, but still informative. 8)
|
:idea: Maybe you Guys should just disable the software n encode a clip to
make sure thats your problem before you go crazy guessing, just left click exit on the icon. |
Kwag
Thought you might be interested in the way that the KVCD method has stopped someone in their tracks 8O Check this 8) http://forum.doom9.org/showthread.ph...0&pagenumber=2 ozjeff99 |
He already knows, I've started hanging out here.
Right now I'm trying to figure out what noise filtering I do pre-KVCD and how that might mess things up. I doubt SpotRemover as a way to fill dropouts and scratches would interfere. Things like DarkSmoother need to be tested. I'm working on field-oriented filtering right now and have run into a couple of problems with the optimal script. AviSynth doesn't have a chr function so there's got to be more script. |
Quote:
Spatial filters work on a frame by frame basis, not on adjacent frames like temporal filters. So maybe a mix of spatial filters will do the job :idea: -kwag |
It's not as simple as that with old film or any analog tape. Look at the image at the bottom of this page: http://www.geocities.com/fredthompson6
This shows a number of things including: 1) scratch/dropout restoration must be done on a per-field basis. 2) spatial filtering must be done on a per-field basis 3) scratch/dropout filtering must be done before decombing (believe me, it's a LOT harder otherwise) 4) chroma filtering needs to be done first In this case, greyscale() or tweak() works for chroma but only because this is a B&W film captured from VHS tape. I'm also looking at near-black swim on camcorder and satellite source. That's where DarkSmoother comes in. Am also wondering how to get rid of the swimming near the edges of satellite MPEG and at the beginning of a scene change. Some of that source is pre-compressed so very good but hte streamed stuff is noisy as all getout. Tried encapsulating the optimal script but AviSynth is choking on it. The idea is to have global variable at the top of the script so Joe Sixpack can easily configure it. These are followed by media-specific cleaning then the field split like so: GetParity() ? nop() : ReverseFieldDominance() even=SelectEven(SeparateFields()).KVCD_Filter() odd=SelectOdd(SeparateFields()).KVCD_Filter() Interleave(even,odd).Weave() (That first line is to ensure order because DV is reversed.) Problem is, AviSynth doesn't like the encapsulation. I'm PMing you the script right now. |
Script just updated.
Now TemporalSoften "radius" paramer is also adaptive ;) Artifacts (Gibbs) are now buried deeper :) -kwag |
hey,
8O few "seconds" out and tons news and tons posts to read! :wink: |
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.