AnimeIVTC() - All-in-one solution for interlacing and blending - v2.00 [2010-01-07]

[Nightshiver]

Rather odd source. Most methods work, but keep some sort of chroma blend/ghost. What worked best for me was

Yadif(mode=1, order=-1)
SRestore()

And then you'll have to manually use FreezeFrame(bad frame#,bad frame#, good frame#) to get rid of the odd blend. Not even VFR worked.

[Emulgator]

Just found a small variable declaration mismatch, easy to fix:
aatype is declared as int, the html mentions int as well , but in the actual 2.00 script a string is expected.
Just open AnimeIVTC 2.00.avsi and change line 14:
, float "dark", int "thin", int "sharp", int "smooth", bool "stabilize", int "tradius", int "aapel", int "aaov", int "aablk", int "aatype"){
into
, float "dark", int "thin", int "sharp", int "smooth", bool "stabilize", int "tradius", int "aapel", int "aaov", int "aablk", string "aatype"){

[aegisofrime]

I'm wondering, could somebody mirror the script and guide on a file hosting service that doesn't suck? I have tried Sendspace for one whole day now, and it won't let me download. Probably because every ISP in my country uses a transparent proxy for their customers. A mediafire mirror would be much appreciated. Thanks!

[Nightshiver]

Here: http://www.mediafire.com/?3ct9wxzdjacj8xw

[aegisofrime]

Here: http://www.mediafire.com/?3ct9wxzdjacj8xw

Thank you very much! That was a lot less painful than sendspace!

[RsVolcom]

Here: http://www.mediafire.com/?3ct9wxzdjacj8xwThanks for this. FYI, It is smaller than the sendspace link, which would not extract.

[Nightshiver]

That's because the sendspace link contains the TGMCMod.

[Shinji-kun]

Hello thetoof

Thread here (http://forum.doom9.org/showthread.php?t=157051). I am trying to restore a berserk trailer that AnimeWorks made (http://www.mediafire.com/?ionx747yy9ch4k0). It has been suggested that the telecine process of the trailer is somewhat irregular. So I tried using your AnimeIVTC() function.

I said it in that thread but, what mode should I use for the trailer? I think it's either 2 or 3 but I still see interlaced frames ("after image" effect in certain frames, don't know if that is interlacing or blending) using both modes. Could you take a look at the sample and give me some help with which mode to use and which parameters to use that might fix the interlacing?

Thanks

[Nightshiver]

The error is not with AAA. "Run out of memory" Well, what do you think it is? It's that Avisynth ran out of processing memory. Use Setmemorymax(number) to correct this. Hopefully you have at least 1gig of RAM in your system, ideally more. (This is for your other post)

It's field-blended. Sometimes there is a 3:2 pattern, other times, there isn't any pattern. You are also using omode and credconv wrong. You don't use those 2 functions with mode=2 or 3. To help you in the future of determining what the source is, I suggest downloading the "old" faq in the original package.

A simple Animeivtc(mode=2,aa=0) works just fine for me on the sample provided. I also would not use AAA or the other filters you currently have chained to this.

Animeivtc(mode=2,aa=0)
source = last
vectors = source.MVAnalyseMulti(refframes=2, pel=2, overlap=4, sharp=1, idx=1)
source.MVDegrainMulti(vectors,thSAD=600,idx=1)
LimitedSharpenFaster(ss_x=1.8, ss_y=1.8, smode=4, strength=180, overshoot=1, undershoot=1)

Works well, and if you want a little more sharpness/thinning, use aa=4.

[Shinji-kun]

which MVTools version did you use?

[Nightshiver]

This one. (http://www.mediafire.com/?1q5yxmmdjqrbklq)

[Shinji-kun]

This one. (http://www.mediafire.com/?1q5yxmmdjqrbklq)

Never mind. 1.9.7. works.

Also, there was still substantial aliasing, grain and banding in the video when I loaded the script you suggested. I think I will use AAA and the others.

I saw interlacing as well.

[Nightshiver]

You claim substantial aliasing,grain and banding while I did not. What are you using to view the output and are you using the sample you provided? There is no such thing as substantial grain after what I did to it, and there is definitely no banding even with grain removal. And don't use the forum for pictures, it's to damn slow. Just upload to tinypic and post a link. I saw no interlacing myself, so please give an exact frame number. You also have dot crawl on the sides, so Checkmate() is needed.

Pics for proof:


Your version: Click (http://img836.imageshack.us/img836/4186/berserk822a.png)
My version: Click (http://img718.imageshack.us/img718/7543/berserk822b.png)
Your version with AAA: Click (http://img715.imageshack.us/img715/690/berserk822awithaaa.png)
My version with LSF: Click (http://img693.imageshack.us/img693/6339/berserk822bwithlsf.png)
Original: Click (http://img269.imageshack.us/img269/9042/berserk822c.png)

Although AAA isn't even a sharpening filter, so it can't really be compared to LSF.

[Shinji-kun]

The images you posted only deal with the grain. There is still aliasing and some interlaced frames when I use your script.

Aliasing

Source
http://img834.imageshack.us/img834/6778/bersrktraileraliasingso.png


Animeivtc(mode=2,aa=0)
source = last
vectors = source.MVAnalyseMulti(refframes=2, pel=2, overlap=4, sharp=1, idx=1)
source.MVDegrainMulti(vectors,thSAD=600,idx=1)
LimitedSharpenFaster(ss_x=1.8, ss_y=1.8, smode=4,strength=180, overshoot=1, undershoot=1)

http://img408.imageshack.us/img408/6086/berserktrailersuggestio.png


AAA()

http://img529.imageshack.us/img529/5968/berserktraileraaagradfu.png (http://img529.imageshack.us/i/berserktraileraaagradfu.png/)


Interlacing

Source

http://img714.imageshack.us/img714/2224/berserktrailerinterlace.png (http://img714.imageshack.us/i/berserktrailerinterlace.png/)


AnimeIVTC(mode=2,aa=0)

http://img689.imageshack.us/img689/1200/berserktraileranimeivtc.png (http://img689.imageshack.us/i/berserktraileranimeivtc.png/)

Uploaded with ImageShack.us (http://imageshack.us)


I viewed these script in vdub 1.9.10.

[Nightshiver]

Give me frame numbers please. And the aa= function of animeivtc is what that's for, aa. I liked aa=2 or 4 for this particular source.

EDIT: Actually, after finding the frame's in question and after testing and thinking about it, those are the only frames that are like that. There are no other frames like that throughout the video, which leads me to believe that what this small part (there are a few other parts) is actually not blended, but it is a video effect the producers/animators put in. To support this, Gut's face is getting an extreme close up at high speed, and in movies and some animation, there are actually blurring/blending effects that are put in to make it look like it's going even faster or to just enhance the effect. I believe that is what the case is here.

Why? Because after thoroughly testing all options of AnimeIVTC, TMPGMCbeta2 and mod, yadif, Srestore, and VFR, those same sections are always like that. I think that this "trailer" is just a special case for this. Where is the source from this? The DVD set from a couple years ago or the "Remastered" set, which is more recent?

[Shinji-kun]

Where is the source from this? The DVD set from a couple years ago or the "Remastered" set, which is more recent?

Giant Robo: Complete Collection. The Remastered is more recent. It's easily obtainable info.

What about the anti-aliasing? I tried the various aa functions in AnimeIVTC and none worked as good as AAA(). The AnimeIVTC aa options always left at least some form of jagged edges in the red parts around frame 1272 (filtered).

[Nightshiver]

Well, if AAA works better for you, then use it. Just out of curiosity, why do you just want the trailer encoded?

[Shinji-kun]

Could you give a link to the old AnimeIVTC guide you mentioned?

[Nightshiver]

Click this. (http://www.mediafire.com/?cnecwea5vm643z1)

[Shinji-kun]

The encode failed again. It gave some error about not being able to cache enough frames and stopped encoding when there was 40 seconds left of the video. This only seems to happen when I use AAA().

About the anti-aliasing, would aa=2 or 4 remove the aliasing in the final encode even if it does not look satisfactory when previewing the script? I think there might be a problem with my video card or MPC-HC because no matter what aa filters I use I can still see aliasing in the encoded X264 MKV files when I view them in mmpc-hc.

I also have a question about an Ed Edd n Eddy episode but I'll post a sample later.

[MatLz]

Hum...I can't see aliasing on the pics...
There is a chroma problem on the blood edges
r=1 #or 2, but not more
deen("a2d",r,1,30,1,1,1)should work.

[Vitaliy Gorbatenko]

Fix chroma problem:
MergeChroma(aWarpSharp2(chroma=4).RemoveGrain(12))

[Shinji-kun]

^ Seems to work. As I mentioned before I have also had a similar problem with Ed Edd N Eddy footage, as well as mpc-hc. The jagged chroma edges are NOT visible when I view the E, E n E episode avs script (or even the source .VOB) in virtualdub, however when I try to view the same avs script in mpc-hc the jagged edges of red/yellow objects ARE visible. This includes when I view the encoded X264/MKV files in mpc-hc, even when using overlay mixer as the renderer, which doesn't even convert the colorspace from YUV IIRC.

So why, I ask, does this chroma problem keep occurring and how can I fix it completely so I won't keep having to add MergeChroma commands to every avs script?

[Nightshiver]

I use AvsP for previewing scripts and VLC to play my files. Did you try the mergechroma on your other sources? You might have to use it everytime, because it might be that each source has a different chroma problem. With mpc-hc, it might be a problem with some of the ffdshow filters. What version of ffdshow are you using? Download the latest one from here (http://sourceforge.net/projects/ffdshow-tryout/files/SVN%20builds%20by%20clsid/generic%20builds/ffdshow_rev3572_20100913_clsid.exe/download). Also, I would suggest downloading CCCP for using things with MPC. So download that from here (http://www.cccp-project.net/) and when you go to install it, un check the box where it wants to install ffdshow, as the version with CCCP is old. And use the MPC version that is included with CCCP, not yours.

[Shinji-kun]

I use AvsP for previewing scripts and VLC to play my files. Did you try the mergechroma on your other sources? You might have to use it everytime, because it might be that each source has a different chroma problem. With mpc-hc, it might be a problem with some of the ffdshow filters. What version of ffdshow are you using? Download the latest one from here (http://sourceforge.net/projects/ffdshow-tryout/files/SVN%20builds%20by%20clsid/generic%20builds/ffdshow_rev3572_20100913_clsid.exe/download). Also, I would suggest downloading CCCP for using things with MPC. So download that from here (http://www.cccp-project.net/) and when you go to install it, un check the box where it wants to install ffdshow, as the version with CCCP is old. And use the MPC version that is included with CCCP, not yours.

I do have the CCCP installed and am using their version of MPC. Although I do have the ffdshow that comes with CCCP checked.

Another problem. I encoded with this:

setmemorymax(2000)
DGDecode_mpeg2source("C:\Users\dvd rip\Berserk Trailer.d2v")
LoadPlugin("C:\program files (x86)\avisynth 2.5\plugins\removegrain.dll")
LoadPlugin("C:\program files (x86)\avisynth 2.5\plugins\yadifmod.dll")
LoadPlugin("C:\program files (x86)\avisynth 2.5\plugins\mvtools.dll")
Checkmate()
AnimeIVTC(mode=2,aa=4)
source = last
vectors = source.MVAnalyseMulti(refframes=2, pel=2, overlap=4, sharp=1, idx=1)
source.MVDegrainMulti(vectors,thSAD=600,idx=1)
LimitedSharpenFaster(ss_x=1.8, ss_y=1.8, smode=4, strength=180, overshoot=1, undershoot=1)
gradfunkmirror()
MergeChroma(aWarpSharp2(chroma=4).RemoveGrain(12))

and I recieved this error:

http://img97.imageshack.us/img97/5165/grab01846.jpg (http://img97.imageshack.us/i/grab01846.jpg/)

What is the problem?

and when you go to install it, un check the box where it wants to install ffdshow, as the version with CCCP is old.

Is there any way to do this without having to reinstall CCCP?

[Nightshiver]

I told you before, you need to use Setmemorymax(). Avisynth is running out of memory to use. How much onboard ram do you have? Divide the total amount of memory (in MB) by 4 and enter that number in-between the brackets. If you only have 1gb of RAM, you have problems, and we'll have to remove some functions/change them.

As for removing ffdshow, all you should have to do is go to the Add/Remove Programs in Windows and just remove it.

[Usedocne]

Filtering in multiply stages via HuffYUV might be helpful too, if he's maxing out his memory?

e.g.
VOB(s) -> First half of filters -> HuffYUV
HuffYUV -> Second half of filters -> HuffYUV

[Shinji-kun]

I have an Ed Edd n Eddy clip that I think might be DHT. There is not constant pattern of 3 progressive frames followed by 2 interlaced. Sometimes there are no interlaced frames for hundrends of frames and other times there will suddenly be a 1:5 pattern. The second thing I see starts when you first see the lamp in Edd's room. The frames look interlaced (fields are present) but there is no combing. I wonder if this is a poor telecine or just a motin=on effect that the animators added. It is clear though that it looks like interlacing even though it is not constant. What is this problem? What AnimeIVTC() mode should I use?

Sample 1 (http://www.mediafire.com/?ahhimxkzf8cumm6)

Sample 2: http://www.mediafire.com/?q999whi40kazmkw

Sample 1 has the part I mentioned above (starts with the lamp) where it looks like there is interlacing but may be an intentional motion enhancing effect originally drawn into the cels.

[Nightshiver]

Another odd source with strange blend-like frames. Mode=3 (DHT) does work, but keeps the strange ghosting images. Even with VFR and trying SRestore, they are still there. If you want more luck, try asking MoMonster in his thread about SRestore.

[Shinji-kun]

Couldn't the ghosting just be an intentionally added effect that the animators put in the original production? Like the Berserk zooming effect.

[Nightshiver]

It could be. E,E, and E would be a little notorious for that, as the creators intentionally made the cartoon pencil work, strange colors, shimmers, etc. So yes, it could just be an added effect.

[Shinji-kun]

Can someone confirm if that's the case?

[egrimisu]

why encodes freezes at some point but only on some sources when ifade=true? using megui(huffy lossles) freeze at some point, using x264 at another point.does anybody else encoutered this problem?
what is the latest version with fixes of animeivtc?might that be the problem...

[Nightshiver]

Read through the thread. The ifade function is broken and there is no fix because toof isn't around anymore.

[egrimisu]

Read through the thread. The ifade function is broken and there is no fix because toof isn't around anymore.

damn,
can i run smartfade manualy(outside animeivtc) i read something that i must run it befor telecide and decimate, how can this be done?


strange i've benn encoding 50 eps from an anime in the past mounth and i had this problem at only one ep. I update mvtools and masktools meanwhile, is smartfade dependent on them or other scripts?

[Nightshiver]

Yes. Smartfade relies on the function daa() within AnimeIVTC and daa() relies on MVtools and masktools.

[egrimisu]

Yes. Smartfade relies on the function daa() within AnimeIVTC and daa() relies on MVtools and masktools.

Hi, i tried with diferent version of mvtools but same thing.
After that i realised that in the avisynth plugins folder i had masktools v 2.a23 and v 2.a45, i've deleted the 2.a23 and now it's working.

PS v2.a23 worked with ifade=true when aa=2 (daa) strange...

Do i need in the plugins folder v1 and v2 of masktools or it is ok if i leave only v2 there?? i'm not sure if masktools v2 is compatible over v1 for older functions/scripts. Thanks

[Nightshiver]

I personally keep both versions in my plugin folder.

[kaishakunin]

Can someone please look at this and give me some hints on how can I IVTC it correctly:
http://rapidshare.com/files/428730305/VTS_01_1.demuxed_cut.m2v
(this is just a 3-second cut from the DVD, size=3Mb)
All methods I tried give unsatisfactory results, the decimated frames are damaged.

[ianken]

Have fun killing all the popus and browser hijacks from the download site. Damn. Even popup blocker failed to stop it. Had to task kill Firefox.

[egrimisu]

Since chrfix=ture does not work when mode=2, does anybody know a workaround to use chrfix? i realy like what it does.

[tasslehoff44]

Hi,

I tried to make a good encoding of a japan anime TS hd file.
I use hardcoded subtitle, trim, resize and some restoring filter.
My encoder gui is staxrip.

I have to deal with vfr and i have some dificulty.

Can you help me ?

[Nightshiver]

Post a sample.

[tasslehoff44]

I don't know how to make a ts sample but if I have the right,
this is here: http://elitistfags.oversubs.org/packs/

on #elitistfags@rizon.net

i make a try on one piece 475

[maki]

I have this m2ts file and I want to know if it's a VFR file or what not.

I demuxed the file and loaded to DGAVCindex. I cut the file short since the file itself is like 4G (DGindex by default has ignore pulldown flags, so I left it as it is and demuxed the file giving me a portion of the h.264 file). This file includes a portion of the beginning and OP of the anime.

http://www.megaupload.com/?d=9255TWVZ

[Nightshiver]

tasslehoff44, you need to upload it to somewhere like Megaupload. I don't have IRC and it's not really worth it for me to install it for the sake of grabbing that 1 file.

Maki, I don't know what the heck you're talking about. Your source is from a BluRay, and it's perfectly 1080 progressive. There's nothing you need to do.

[tasslehoff44]

I cut the opening with FameRing Smart Cutter for DV and DVB v1.3.2, tell me if there is another better tools.
Sample (http://www.megaupload.com/?d=AV21MD50)

[Nightshiver]

Tass, your video isn't VFR and it isn't interlaced. While it does show flags of being 29.97, it's actually progressive. You don't need to do any kind of decimating, IVTC, or de-interlacing.

[tasslehoff44]

Aparently, there is no interlaced frame in the opening, but it's realy an hybrid source, see the frame capture (http://img526.imageshack.us/img526/8753/475w0.jpg).
Mediainfo and other tool report bad info about fps, but media info report that is an interlaced frame.

I do some test and it appear that the opening don't have variable frame rate, but the episode have it.
I upload the encoded opening and episode with the timecode but non muxed: Episode (http://www.megaupload.com/?d=RL75LZ8Z), Opening (http://www.megaupload.com/?d=UPMTI7NN)

[Overdrive80]

Well other case that i have to use this filter. Post update filter for which have problems with errors merge windows. ^^