This script still good ?

[Mister Hatt]

Sincresize is possibly the most useless resizing filter unless used for VERY specific analog-digital conversion progressing (seriously, read the sinc function page on Wikipedia, it's interesting) and I don't know why anyone would use it. Spline36resize remains the best filter in there, although bicubic/bilinear still have their uses.

@Heavy amv, just so you know, the reason your halos are so bad is probably because warpsharpening at high depths has that effect on anime. Your filtering is way too strong and ripping the DVDs is significantly faster than all the time you're spending tweaking something broken as shit. If the ghosting is in your mkv as well, it is impossible to fix, so I would suggest just sucking it up and grabbing the discs already rather than wasting everyone's time. This footage is known to be a pain in the ass; pay attention to the advice you've been given already and get the results you're looking for.

[mirkosp]

If the ghosting is in your mkv as well, it is impossible to fix, so I would suggest just sucking it up and grabbing the discs already rather than wasting everyone's time.

Ok i found the dvds today in a second hand shop. I use the script in the other thread and i come back to you if i have some issues...

[Heavy amv]

About the last pack of pictures, i remember you that the original footage is to the right.

About the dvds, the first dvd is just ugly... the others are quite good and not so hard to clean. So i rip it, make d2v files, use the script and edit, in order to save all the details i love so much and for America... God blesses me.

[EvaFan]

I'd ask misterhatt to tackle the My hime R1's but he's prolly angry with me .

I gave up hope on them, well specifically a few discs from what I remember. I think only the first 2-4 discs were a blended nightmare, don't remember anymore.

[Mister Hatt]

Rather than being emo about it, upload a sample as well as your script etc so we can take a look. I like puzzles, so finding odd video patterns and such is fun for me, I like the challenge. Maybe I should make a thread to figure out all the really weird shit people come across. I'll write it after those stickies that will never happen.

[EvaFan]

Sample It's 50MB, dunno how the transfer rate will be since you live far away, hopefully not to bad. It doesn't really have any heavy motion in it but it should be good enough to see how terrible it is.

I haven't really done any work with them since this thread: http://www.animemusicvideos.org/forum/v ... 45&t=85412

That was a really long time ago, back when most of the stuff I knew was from the guides at the time which were even more outdated then they are now. I confess I really still don't know much about how to deal with blending anyway though other than what is there. Blending has been something I've just flat out avoided at all costs up till now. It sucks cause you can't tell if the R1's are bad unless you buy them or rent them. At that point spending more money on R2's seems wasteful.

[Mister Hatt]

Sorry for the delay in replying, took me a while to get the source. Mind posting your script?

I just assume that all R1's suck, IMO the only country that can reliably produce decent DVD masters is Italy, and even then only from one company.

I'm currently looking over your source in YATTA. It is quite strange and while it has some distinct patterns, they either don't make sense or cannot be dealt with in any way I can tell. Every 4th and 11th set of fields in a group of 15 frames has a chroma offset, and there appears to also be a distinctly different pattern between the background clouds and your heroine's face.

Not too sure what is up with it to be honest; other than it being an absolutely terrible master. I'll keep hacking at it but it is well past the point where I would write a letter to the company telling them how shit a job they did, demand a refund, and buy the R2j. If the BD doesn't suck, might be worth grabbing the Japanese BD instead.

[EvaFan]

Code: Select all

MPEG2Source("C:\Hime01.d2v", cpu=4) 
TDeint(mode=1,order=0,mthreshl=8,mthreshc=8,chroma=true,tryweave=true,denoise=true) 
Blendbob(badframes=0) 
Tweak(hue=1.0,sat=1.2,bright=-0.1,cont=1.0) 
Convolution3d(preset="animeHQ") 
RemoveGrain(mode=17) 
warpsharp() 
fastlinedarken(thinning=0) 
crop(0,2,0,0) 
lanczosresize(640,480)

Keep in mind this is from 2007.

That was the last thing I ended up with. I only knew about stuff from the guides back then too. I just ended up trying unblend() and the doublefps+bob route and that's it cause I didn't and still don't know much about dealing with blending. I just avoid sources like this instead of dealing with the hassle. I doubt I'll ever buy the R2's now that I own these shitty DVDs. They are shit but they still play if I or any friends want to watch em. Kinda sux since I can't use them to edit but oh well. I can get you a different sample with more motion if you want, up to you. It's probably not even worth the headaches to find a fix for them.

[Mister Hatt]

I would have preferred a sample with more motion but I know what's up with it now. At first I thought it was a ccncn pattern which was confusing TFM, but on closer inspection that wasn't the case, and TFM can handle it anyway. I'm not too sure what's up with it as far as the coding goes but both TFM and Telecide absolutely refuse to work on it, even though it is clearly 3:2. Next try was with YATTA which died in the ass as well (I have NEVER seen this before.) Eventually by about 3AM I figured out it was split into several onscreen patterns. The luma is split on the background and forground (different patterns, but the offset is easy so you can tweak your decimation to cover the background) while the chroma also has a different pattern.

I figured I could just IVTC by hand and with overrides the luma/chroma and then merge them back together. This was not the case. The individual fields have chroma blending, which is nigh on impossible to remove. I've given up on it now that I know how it works, and how long it will take just to correct the luma, while the chroma is pretty much stuck as is.

Important things to note are that it was telecined, and your script did not address that. Seems a lot of people have trouble figuring out frame coding and field operations. Also, it took me about an hour to do 2000 frames of the luma alone once I knew what I wanted to do, I imagine it would take you a ridiculously long amount of time to do 26 episodes, and while still having broken chroma. Could probably do some hack to clip the chroma based on luma, but it seems a little pointless considering how well it would most likely turn out.

This can be fixed, but I don't see the point. You basically got screwed by the industry not caring about what customers think of quality, and knowing they can get away with it. Uploading my ovr files for great justice etc.

http://pastebin.com/MhU7WihK TFM
http://pastebin.com/mA1i4My2 TDecimate