View Full Version : New? AviSynth VFR VS Trigun (please read and suggest)
OvejaNegra
6th June 2008, 22:15
First:
Please don't mention animeIVTC, is very unstable right now and i can't make it work.
Second:
I'm trying to process trigun right now. The DVDs are an awesome mix:
Telecine fragments
Pure interlaced (hard coded): Pans, zooms, end credits.
Total Telecine opening credits (correctly flagged, no problem here)
My goal
Use TFM to do field matching if possible-> send hints to next filter for the combed frames
Use Tdeint with full=false hints=true for bobing only the interlaced parts. This in theory should give me good bobed credits and zooms, pans, etc. IN THEORY tfm will hint the interlaced/unmatched frames and Tdeint will only touch those.
Finally , apply tdecimate for producing a VFR matroska with 23.97 and 59.94 sections.
In theory...........
This shold work:
tfm (order=1,mode=5,pp=1,display=false)
tdeint(mode=1,order=1,field=1,full=true,tryWeave=false,hints=true)
But tfm hints the credits and the Zooms,Pans as Clean Frame!!
Maybe because it's harn interlaced encoded over telecine.
Second approach:
tfm (order=1,mode=5,pp=1,display=true)
tdeint(mode=1,order=1,field=1,full=false,tryWeave=false,hints=false)
(Forget the hints and let tdeint detect and process the interlaced frames, this maybe gives me accidental bobbing on some progressive frames but i have no choice.)
this does not works. Maybe because tfm change the order of the fields. The credits jumps up and down som times and the pans are a little............... :)
Sugestions?
Advices?
Is there any other way of doing this?
Or this will never work?
PD: i'm not sure about the settings for tdecimate, please anybody post some example.
Thanks.
TheRyuu
8th June 2008, 04:03
It would be incredibly easy to do it in yatta with sectioned filtering but that might be a little over your head.
I wouldn't bob the interlaced parts either, I'd try and keep them at 29.97fps but that's my personal opinion on it.
OvejaNegra
9th June 2008, 14:04
Well, i can't find a simple tutorial for yatta. I know how tivtc works, but i don't understand how apply all that in yatta.
I wan't do this auto, that's all. It's not so simple as credits - episode - credits (i do that by hand) the episodes have a LOT of panning scenes with hard interlacing.
The bobing is for those scenes, now they look smooth. The credits need it too.
DarkT
9th June 2008, 17:50
Huh? Waht you talking about? My team did trigun from messed up USA sources, and we did it fine, what I did at the end, as the green lyrics are hardsubbed... I simply made it a fake 16:9 :), I cropped it to hell, and then added black borders :).
I duno what you haveing problem with.. It's true, each dvd, there's 1 episode which needs a different TIVTC line processing it, since it becomes screwed up - like that episode when there's that scene with the well...
But generally it's easy...
Here's the script I use for the USA dvds:
d2vpath = "VTS_01_1.d2v"
MPEG2Source(d2vpath,info=3)
Bifrost(Interlaced=true, scenelumathresh=5)
TComb(mode=0,fthreshL=5,othreshL=7)
DeGrainMedian(LimitY=3,LimitUV=4,Mode=4,norow=true,interlaced=true)
VagueDenoiser(threshold=0.5, method=3, nsteps=5, chromaT=2.0, interlaced=true)
interp = AssumeTFF().separatefields().SelectEven().eedi2(field=-1,pp=3)
deint = TDeint(mode=0,Order=-1,type=1,AP=55,edeint=interp,slow=2)
TFM(d2v=d2vpath,clip2=deint,order=-1,Mode=5).TDecimate(mode=1, denoise=true)
DFMDeRainbow()
Levels(16, 1, 255, 0, 255, coring=false)
Tweak(cont=1.00, sat=1.0,bright=3)
crop(12, 4, -12, -0)
BilinearResize(768,576)
RemoveGrain(mode=1)
Convolution3d (preset="animeHQ")
hqdn3d(ls=1.2,cs=1.0,lt=1.2,ct=1.0)
DeHalo_alpha()
Toon(strength=0.5)
WarpSharp(96, 10, 128, -0.6)
AA()
LimitedSharpenFaster(Smode=4, strength=77)
deblock(quant=18)
Trim1 = Trim(0,111)
Trim2 = Trim(112,232).crop(118,0,-130,-0).Addborders(118,0,130,0)
Trim3 = Trim(233,32481)
Trim4 = Trim(32482,34283).crop(0,26,0,-152).Addborders(0,88,0,90)
Trim5 = Trim(34284,34669).crop(0,126,-0,-52).Addborders(0,88,0,90)
Trim6 = Trim(34670,35412)
Full = Trim1+Trim2+Trim3+trim4+trim5+trim6
FullTrimmed = Full.Trim(2151,35412)
Return FullTrimmed
that's 1st pass(lagarith loseless), episode 10, 2nd pass(auto 2-pass h264 in megui):
Trim1 = Trim(0,32692)
Trim2 = Trim(32693,34283).DFMDERAINBOW(7).DFMDERAINBOW(3)
Trim3 = Trim(34284,35412)
Full = Trim1+Trim2+Trim3
FullBlack = ImageSource("BLACKNESS.PNG", start=1, end=9,fps=23.976023976023976023976023976024).Converttoyv12()+Full
FullSubs = FullBlack.TextSub("T10 DUAL VERSION HARDSUB logo+kara FOR release.ass")
Return FullSubs
The ED in the USA dvds is very rainbowed, hence you se ein that section a double dfmderainbow...
Anyway, hope I helped.
Edit:
Just checked, the levels line is messed up, heh... Anyway...
Edit2:
There's sweet R2J's around at asiandvdclub right now, we're gonna get htem to do V2, you might want those as well, as they are much better quality then the US version.
OvejaNegra
9th June 2008, 23:04
Im looking to your script right now, thanks.
But please note something:
I'm trying to preserve the bobing in the panning and zooming scenes, look a lot better (yes, i know, i'm a fan of the 60fps videos, since i watched a pron processed with dgbob, the movement was soooooo gooood).
My goal is preserve those scenes, if everything was just create 23.97 i wont be creating such mess.
PD: carefull with the forum rules please
DarkT
9th June 2008, 23:20
Hmmm, I dunno man, I don't remember any problems with Trigun - but then again, I hate that project and only do it since the team voted it in >.< so I don't really check it thoroughly - could have missed the zooms/pans/whatever that you're talking about...
OvejaNegra
9th June 2008, 23:58
Forgive my PAL land ignorance but:
Why everybody is so interested in the region 2 DVDS? Because its 24 fps accelerated to 25 without telecine?
thanks
martino
10th June 2008, 00:20
Note: USA is not PAL. USA is NTSC R1. NTSC (R1) == NTSC (R2), you'll still have the telecine there, most likely anyway. *wonders whether there's been a progressive R1 and telecined R2 or other way round*
Also R2s are usually better sources. Why? Who knows... NA companies sucking at mastering? Perhaps...
P.S. Auto is not always good, and not always possible with certain sources. If you REALLY want to do a good job take a look at YATTA as Dragon suggested. Simple manual, extensive manual... no manual, if you really want you can get around it. All you need are nerves (with that should be included a working brain, rite), coffee and painkillers.
OvejaNegra
10th June 2008, 00:25
Wel ,since R2 is Europe i assumed R2=PAL, sorry.
It's just better sources and transfer?
Ok, thanks for the info.
martino
10th June 2008, 00:33
Well, I always wondered about the naming too. Europe is PAL, marked as DVD region 2. But note that NTSC region 2 is Japan (anything else there?). I get confused myself at times. Anyway, if someone has a better explanation feel free to input it, or correct w/e that's been typed in this thread.
And just to conclude it, stay off anime PALs.
P.S. Sorry for putting this thread offtopic.
DarkT
10th June 2008, 00:34
Don't touch YATTA - once you go YATTA you never go back ;).
As for why R2J are better then their US counterparts - there are like 2 main reasons: 1)US usually will try and squeeze more episodes into a dvd and 2)they mess shit up in their "transffers" (or whatever it's called).
A very known example is Full Metal Panic? Fumoffu, which is just HORRRRIBBBLLEE in the US version - it has tons of rainbowing and tons of dot-crawl, and is less sharp, and what not... the R2J is so crystal clear and moeeeee, it's oishi :).
Anyway...
Everybody are interested in the R2J (the Japanese R2) since they are like "1st source" and then you have US - which is like "2nd hand source" - and THENNNN you have the PAL anime dvds - which nobody works with to my knowledge - but I have heared some blending horror stories about those :).
thetoof
12th June 2008, 07:13
Please don't mention animeIVTC, is very unstable right now and i can't make it work.
Yeah, that was cuz you used an inappropriate mode. Mode=7 is indeed experimental and is not intended for ivtcing, only for compressibility improvement through vfr. Try the proper mode (3 or 4) for your problem before claiming it's unstable. ;)
Pure interlaced (hard coded): Pans, zooms, There is a very high probability that this is double hard telecine (mode=4 for your case). I'd need a sample to confirm this though...
end credits. and this is most likely 30i (truly interlaced) on top of 24t (telecined)
Comatose
14th June 2008, 09:30
Speaking of samples, could you please provide a sample for each type of thing-that-needs-to-be-processed (e.g. double hard telecine) in the AnimeIVTC thread?
AnimeIVTC requires you to be able to tell them apart but it's not that obvious for someone who's never really looked into it.
thetoof
15th June 2008, 14:37
There are screenshots in the first post and in the guide + an explanation to know which type of source you have by using DGIndex and your eyes... if you still need some video samples, I could still post some.
OvejaNegra
16th June 2008, 01:24
thetoof:
Sorry, my statement was not a critic. Since i saw some people reporting problems in the main trhead i assumed it was the normal-first-release-bugs.
AnimeIVTC is very stable (exept for mode 7 :)
"only for compressibility improvement through vfr"
That was axactly what i was trying to do.
"There is a very high probability that this is double hard telecine (mode=4 for your case). I'd need a sample to confirm this though"
I will create on tonight
"and this is most likely 30i (truly interlaced) on top of 24t (telecined)"
yes, the credits are interlaced (for smoother scrolling) they look nice with a simple bob().
thetoof
16th June 2008, 04:51
Yeah, I'll add decimate=6 soon (vfr decimation for any type of source, not only hybrid) to replace the buggy mode=7. About the sample, it'd be nice that you post it in the animeivtc thread, since it may help many understand what dht is.
vBulletin® v3.8.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.