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

[thetoof]

Somewhere, you must have applied blend-deinterlacing to your sample.
What is the workflow you used to end up with the file you uploaded (capture, selection, extraction....) Please mention the programs & settings used.
BTW, d2v is the text file associated with a mpeg-2 stream, while m2v contains the actual video stream.

[Shiznadnaz]

The first sample I uploaded, I used DGIndex to demux the video, the second one I used VideoRedo to split up the file. I made a d2v using DGIndex using the default settings and used AvsP to view the d2v and found that image I uploaded.

[thetoof]

Honestly... I don't understand how you can see combed frames and upload the very same file.... that is simply blended.
Try running AnimeIVTC(mode=1) instead of srestore. If your source is telecined (i.e. combed frames), it'll work... if it outputs some blends, then you need srestore. A safe way of processing your source if it has both (very first time I'd hear of a case like that), apply srestore without decimation and then run it through animeivtc, which will catch any remaining combing and apply decimation without harming clean frames.

[Shiznadnaz]

I have just noticed something, there are no combed frames in the actual video but when I view the d2v file (created in DGIndex) in AvsP there are combed frames for some reason :confused: I must be doing something wrong. I got the images from the d2v file. Could you show a little example of how I would use AnimeIVTC with srestore?

[thetoof]

This.... is very weird.

Try encoding a lossless rendering pass in vdub with this simple script:
mpeg2source("yourfile")
srestore(omode="pp3")
You should not see anymore blends in the resulting avi file.

Please try and report

A usage of srestore + animeivtc would be to simply add animeivtc(mode=1, other settings) after the above script, but please test the script before. I don't want to know what the preview is, but the actual file after processing.

[Shiznadnaz]

Okay, srestore removed the blends but I can't use srestore and animeivtc at the same time or in the same script, when I try to use them both vdub crashes and closes down. Would it be alright to use animeivtc on the lossless avi file without blends?

[thetoof]

Yes, if there is still some combing left.

If it's clear, load it, (t)decimate() and add the rest of your filters.

[Shiznadnaz]

I just found out where those combed frames came from. After DGIndex finished making the d2v file, a window popped up saying, http://i34.tinypic.com/qn7gva.jpg and I obviously did not know what this meant so I clicked on yes. This was the file I got the image with combed frames from but later on I decided to check the file which was replaced and it only had blends. So I guess I only need to use srestore now and it'll be much faster, sorry for all the trouble and thanks for your help.

[Rarok]

Hello. I'm very, very, very noob at this so please don't be too hard with me.
I'm having a lot of troubles with a 1080i .ts anime raw, Dgindex says that is NTSC and I see a 3:2 pattern but the Opening & Ending are also 3:2 so: is this a hard telecined or what?

[patrick_]

If it has a 3:2 pattern all over, you don't need AnimeIVTC. You'll be fine (or actually better of) with IVTC()

[Nightshiver]

Upload a sample

[Rarok]

Ok, hero goes a sample http://www.megaupload.com/es/?d=36PXFQ46 , it's quite bigger but I tried to take vertical and horizontal movement. It's a bit from opening but all the chapter has the same type of telecined.


I tried using a standard IVTC with

LoadPlugin("DGDecode.dll")
#LoadPlugin("TIVTC.dll")
video=MPEG2Source("RAW1.d2v")
audio=DirectShowSource("RAW1 PID 141 DELAY -120ms.mka")
AudioDub(video,audio)
tfm(d2v="RAW1.d2v")
tdecimate(mode=1)

But I'm having many failures in the image, you can see some snapshots:
http://hdimage.org/images/mqxk6ahwt98stihiac6_851_thumb.png (http://hdimage.org/viewer.php?file=mqxk6ahwt98stihiac6_851.png)http://hdimage.org/images/lnmi6qvhuo13o3vrjo_6390_thumb.png (http://hdimage.org/viewer.php?file=lnmi6qvhuo13o3vrjo_6390.png)http://hdimage.org/images/nxfqslj8w1h5864alt19_27168_thumb.png (http://hdimage.org/viewer.php?file=nxfqslj8w1h5864alt19_27168.png)
As you can see the first file has some strange horizontal lines and the other two have some kind of duplicated imagen in the upper side of the objects, I don't know if it's because a bad IVTC or if the original image had they.


Maybe this is easy but I don't know how to handle it, sorry.
Also I'm sorry because my English isn't good

[thetoof]

1 - 314.86MB is way to big. 10-20mb should be more than enough.
2 - What the heck is your source? mpeg video and mka container for the audio? Anyways, the sample should be created using DGIndex - save and demux video with honor pulldown flags

[Nightshiver]

I was about to comment on why it was .avi

[Rarok]

I didn't know how to create the sample so I take it from vdub in lagarith.
The source is mpeg video and audio is aac, I didn't found how to use it with avisynth so I merged it into a mka container and worked fine :(

http://www.megaupload.com/es/?d=W7F5AC07

[thetoof]

It is, as you thought, hard telecine. The artifact you're seeing is called haloing and must be dealt with something else. A few discussions and filters are floating around and here are a few things to play with : mt_convolution, blinddehalo3, dehalo_alpha, yahr...

You could post your problem in this thread, as the title is "general" enough for you to bump in. http://forum.doom9.org/showthread.php?t=143687

Btw, your source is a typical example of badly upscaled material for HD broadcast. You could downscale it to SD without any big loss in quality compared to the compressibility gain. (if you do that, don't forget to use ColorMatrix(mode="Rec.709->Rec.601"))

edit: oh, and the horizontal lines you're talking about = banding. You can dither with gradfun2db and ajust a few settings in your encoder to keep the grain. More info: http://forum.doom9.org/showthread.php?t=143842

[n2k3]

I get this error:
http://i42.tinypic.com/2w5ss8x.png

when I try to run this script:

MPEG2Source("C:\Encoding\SS2\1\vob\VTS_02_1.d2v", cpu=0)
Lanczos4Resize(720, 540)
AnimeIVTC(mode=3,aa=4,omode=1,mix=4,estart1=0,eend1=31416,istart1=0,iend1=0,estart2=33435,eend2=34324,check=3)

I'm trying to get the frame numbers for istart1 and iend1 with check=3
Any player I feed this script crashes. I think its check=3 doing something weird.

Windows XP x64 bit SP-2
Intel Core 2 Duo E6750
Corsair TWIN2X2048-6400C4 (x2) = 4GB RAM

[Nightshiver]

Why are you using media player classic? Use AvsP.

[n2k3]

Why are you using media player classic? Use AvsP.

Like I said, every 'player' crashes, whether its VirtualDubMod / Media Player Classic / AvsP's internal player ;) (I use mplayerc as external AvsP player)
Reason for that if I get this error while using AvsP's internal player the program crashes & I lose my changes made to the script.

[stpdrgstr]

Hi, it's me again, just to report that my old issue was a mistake :p.

And, well, a question is ok too :lol:.

I've been using AnimeIVTC with Serial Experiments Lain DVDs, so far very good results, but, I wonder how much vidthresh should I use for it, because it seems some dups stay in the motion parts. Is an hybrid source with true interlaced credits, and DGIndex reports about 71% Video overall. I wonder if I should put less than 0.6 of vidthresh or more than the default?

[n2k3]

Seems that my problem with check=3 was caused by mvtools with sse3 support (josey_wells' multithreaded branch).
I cannot use SSE4 dll because I only get Illegal Instruction errors
I cannot use SSE3 dll if I want to use check=3.
Now using SSE2 dll, everything works I guess, just slower :(

Question, should I use the original mvtools now (because it does support sse4?) so its faster, or does multithreading still win at SSE2..

[thetoof]

I wonder if I should put less than 0.6 of vidthresh or more than the default?Could you upload a sample of the problematic section?

Question, should I use the original mvtools now (because it does support sse4?) so its faster, or does multithreading still win at SSE2..
Personnally, I prefer using the official v2. Compatibility to come in 1.07 (this should be dealt with over the week-end, as I have a bit more time these days)

[stpdrgstr]

Well, it's not really a problem, there's just some dups left in a section...
But I'll have it uploaded tomorrow. Today I'll be a bit bussy. Maybe at the night :p.

Oh yeah, there's also something weird that has been happening every time I use AnimeIVTC. It's not a problem, nor something that hinders process, but is annoying nonetheless.

Every time I finish the last pass with mode 7, if I just close VirtualDub, the gui closes, but the program is left on background, completely eating one of my both cores (but everything is been finished already). I have to kill the process every time. Happens with every source I've used. Is it my fault, an AnimeIVTC bug or a VD bug?

[MADAJ]

hi...

what is wrong with my script? I have Warpsharp plugin

http://img136.imageshack.us/img136/849/errorvo3.jpg

[Nightshiver]

Perhaps you don't. Do you have the warpsharppackage which includes:

avsfilter.dll
LoadPluginEx.dll
warpsharp.dll

There is a seperate plugin that is called aWarpShapr.dll, but is not the same.

[n2k3]

Personnally, I prefer using the official v2.

How would I make AnimeIVTC work with v2?
As for now if I only put vmtoolsv2.dll in the plugin folder its screaming there is no function named "MVAnalyse" used in the TempGaussMC_beta1mod.avsi, line 167

[Nightshiver]

That's because Fizick made mvtools2, and he did not include MVanalyse in mvtools2, so you can't use it with his. To use Fizicks branch, you have to set MT=false in the animeivtc line.

[Adub]

Don't remove the v1 mvtools. The dll's are named so that they can be used side by side in the avisynth folder. Even the function declarations are different. Leave both mvtools.dll and mvtools2.dll in the avisynth plugins folder.

[Nightshiver]

I never said remove the v1, if your talking to me.

[n2k3]

That's because Fizick made mvtools2, and he did not include MVanalyse in mvtools2, so you can't use it with his. To use Fizicks branch, you have to set MT=false in the animeivtc line.
According to the script the default value of MT already is 'false' (I never put MT=true in my AnimeIVTC line)
Basically you're saying by putting MT=false in there will make use of v2 ?

Don't remove the v1 mvtools. The dll's are named so that they can be used side by side in the avisynth folder. Even the function declarations are different. Leave both mvtools.dll and mvtools2.dll in the avisynth plugins folder.
Because you're saying the function declarations are different how do I tell the script to use v2?
(Also there's no mvtoolsv2.dll with the requirements, so I guess I can't use v2 now with AnimeIVTC?)

I never said remove the v1, if your talking to me.
Nope I wasn't talking to you in particular (as I quoted thetoof).

---
Sorry if its not that obvious for me :(
If I can't use v2 in a simple way (with AnimeIVTC) there's no problem, just say I can't :p

[Nightshiver]

Ah, yes. Going through the required plug-ins folder, I don't think you can use mvtools2 right now, but probably in the next version you will be able to.

[thetoof]

Exactly, for v2 you need to remove all idx, create required super clips and change the MVxx to Mxx. It's all been done and I'm about to finish adding all the new features, so as I said previously, expect v1.07 by the end of the week-end (unless, again, I'm interrupted by something else).

[n2k3]

Exactly, for v2 you need to remove all idx, create required super clips and change the MVxx to Mxx. It's all been done and I'm about to finish adding all the new features, so as I said previously, expect v1.07 by the end of the week-end (unless, again, I'm interrupted by something else).

Ok sure, thanks for the heads up :) awesome job so far!

--- Something else I want to ask
I have a small problem getting these credits (sample) (http://www.myupload.dk/showfile/13893960587.rar) properly displayed. I'm probably doing something wrong..

avs script + comments (also included in the rar)

#--- Source ---
MPEG2Source("C:\Documents and Settings\n2k\Desktop\credits.demuxed.d2v", cpu=0)

#--- IVTC ---
AnimeIVTC(mode=3,aa=4,omode=1,mix=8,pstart1=0,pend1=340,estart1=272,eend1=330,MT=true,overlap=6,pel=4,search=3,check=2)
#~ AnimeIVTC(mode=3,aa=4,omode=1,mix=1,istart1=0,iend1=350,estart1=280,eend1=330,MT=true,overlap=6,pel=4,search=3,check=2)
#Yes I only use 1 at a time -- doh

#--- Crop & Resize ---
Lanczos4Resize(720, 540) # Resizing to fix the non-square pixel aspect ratio

#--- mix=8 --- I think the credits are
#using check=10 for the progressive part, check=2 for the normal interlaced part (in this sample there is none though)
#then removing the check parameter and it says the video is using different frame rates.
#while in check=2 mode the text says the fps is changed to 23.976 while mplayerc says its 19.181 fps?
#using estart1 and eend1 as dummies, because there's no mix mode where you only select a progressive part (for these kind of samples :p)

#--- mix=1
#using check=3 for the true interlaced part, check=2 for the normal interlaced part (in this sample there is none though)
#then removing the check parameter and it says the video is using different frame rates.
#while in check=2 mode the text says the fps is changed to 23.976 while mplayerc says its 19.181 fps?
#using estart1 and eend1 as dummies, because there's no mix mode where you only select a progressive part (for these kind of samples :p)

#--- other comments
#I think the credits are progressive (since they give a better result) but there's still some crawling blurryness.
#Is this fixable with AnimeIVTC, if so how? :) -- The sample screenshot of double hard telecine fix was so nice :'(

[thetoof]

Your credits are interlaced. A very good explanation of why they can't be displayed properly, even after uber-HQ bobbing can be found here : http://forum.doom9.org/showthread.php?p=1185370#post1185370

I don't really get your i/estart/end settings... but instead of going deeper in any explanation, could you wait for the next release where this aspect of aivtc will be much simpler?

[n2k3]

I don't really get your i/estart/end settings... but instead of going deeper in any explanation, could you wait for the next release where this aspect of aivtc will be much simpler?

I can probably wait till the next version ;)
To explain my way of obtaining the frame numbers, so I don't make the same mistake even with the new version :p.
From the documentation:

e1/e2/e3 = sections of the episode that will be IVTCed/converted from a region to another
i1/i2 = sections containing truly interlaced credits - I was assuming the credits might be truly interlaced
Mix=1 : i1 + e1 - I was using mix=1 because there was no 'truly interlaced only mode' - you could drop the e1 values
i1 (istart1/iend1) - Use check=1 for omode=2 and check=3 for omode=1

I'm using omode=1, so using check=3 to get the frame number for the truly interlaced part estart1=0 (start of the clip)
until a reasonable point so the script can combine the 'normal interlaced part' which wouldn't be necessary for this sample a random value I chose was 350.

Since I have to use (according to the documentation) different check modes (check=3 for the truly interlaced parts, check=2 for the normal interlaced parts) I copied the time in mplayerc from frame 350 in check=3 mode, pasted the time in check=2 mode and gotten the frame 280 (like in the comment in the script, somehow the fps changed to 19.181...)
and the end of the clip in 19.181 fps is frame 330.

---
I hope this all clear, and get what I was doing.
Please correct me where I'm going wrong :)

A non related question, where can I find out the difference between hard telecine and double hard telecine? :B

[stpdrgstr]

Hi, I've tested again and the dups overall go away if I get the vidthresh up, as it was to be expected. But then, what does it mean if DGIndex shows 71% video?

[thetoof]

(like in the comment in the script, somehow the fps changed to 19.181...)
and the end of the clip in 19.181 fps is frame 330.
This, I don't get it... let's wait for the next v to see if it has been corrected.
I hope this all clear, and get what I was doing.
Please correct me where I'm going wrong :)
Oh, do you mean you simply wanted to deinterlace the credits and convert them to 23.976? If that's the case, may I ask why?

A non related question, where can I find out the difference between hard telecine and double hard telecine? :BNext v's docs will be much better with a lot more images and examples to explain everything. Basically, HT = 3:2 pattern and DHT = apparently no pattern, but when you look at the front and the background separately, you'll see a 3:2 pattern in both, though they are not synchronized. http://forum.doom9.org/showthread.php?p=1221663#post1221663

Hi, I've tested again and the dups overall go away if I get the vidthresh up, as it was to be expected. But then, what does it mean if DGIndex shows 71% video?Good to hear. It only means that 29% of your clip has soft pulldown... the rest can be hard pulldown, 30p or 30i.
About the bug you mentioned, I'll have a look into it (+ tell me if next v corrects it)

[n2k3]

This, I don't get it... let's wait for the next v to see if it has been corrected.
Yup tried it again, and confirmed it with the internal player of AvsP and vdubmod, they all say it ended up using 19.181 in check=2 mode.


Oh, do you mean you simply wanted to deinterlace the credits and convert them to 23.976? If that's the case, may I ask why?
Basically yes, why you ask? Because I thought I could get rid of t he crawling blurry parts of the text in the credits using the magic of aivtc :)
(using mode=1 doesn't get me the nice clean scrolling credits) - pic (http://i43.tinypic.com/8xk6wy.jpg)

What I want is having smooth panning credits at 23.976 fps without the blurry parts on some of the text, so I have a cfr clip

Next v's docs will be much better with a lot more images and examples to explain everything. Basically, HT = 3:2 pattern and DHT = apparently no pattern, but when you look at the front and the background separately, you'll see a 3:2 pattern in both, though they are not synchronized. http://forum.doom9.org/showthread.php?p=1221663#post1221663
Thanks for the explanation :)

[Leak]

Yup tried it again, and confirmed it with the internal player of AvsP and vdubmod, they all say it ended up using 19.181 in check=2 mode.
19.1808 is the framerate you'd get if you apply TDecimate twice, i.e. 29.97*(4/5)*(4/5)...

[thetoof]

Basically yes, why you ask? Because I thought I could get rid of t he crawling blurry parts of the text in the credits using the magic of aivtc :)
(using mode=1 doesn't get me the nice clean scrolling credits) - pic (http://i43.tinypic.com/8xk6wy.jpg)
What I want is having smooth panning credits at 23.976 fps without the blurry parts on some of the text, so I have a cfr clipIt's actually the "magic" of tempgaussmc you're seeing here. AnimeIVTC is meant to simplify the identification of the problems and the converting/decimating, trimming and splicing process, which is why there is no "interlaced only" support since it's basically source.tgmc.mvflowfps (can easily be done externally). If you want to use animeivtc, load the opening + episode + ending in DGIndex to create a single d2v file (so a single clip to process with mix=1)
As Didée explained in the post I linked you too, the data you're trying to recover doesn't exist due to the scrolling speed or the interlaced elements (i.e. some horizontal edges can't be seen/recovered)

19.1808 is the framerate you'd get if you apply TDecimate twice, i.e. 29.97*(4/5)*(4/5)...Oh, you're right... I must've done something fuzzy. *will fix*

[n2k3]

Just came to the idea of fixing the credits using a creditless source and overlay the credits with ass styled text.

[thetoof]

Sure you can... but that's doing it the hard way ;) Totally doable though.

[vanger]

Hi!
I tried this script on Robotech Remastered source. It is NTCS 29.97fps with Double Hard Telecine.
I used string
AnimeIVTC (mode=2,aa=2,MT=true)
and have speed 0.5fps compressing to HuffYUV and 25-30% CPU load (I have Phenom X4 2.2GHz).
Any ideas how to speedup?

[thetoof]

For now... lower precision.

[vanger]

But how to load cpu closer to 90-100% ?

I have an idea to start 3 processes at once with different episodes :)

[zocz]

Hi Everyone!

I would like to encoding some anime, and I have a problem.
I cannot identify the source (ie: soft telecine, hard telecine, double hard telecine etc.)
How can I do that? What program or what kind of method could be useful to do it?
Would be important to know because of AnimeIVTC settings.
Thanks in advance!

PS: My English is poor, so be patient.

[Nightshiver]

You can make a separate thread and post a sample of untouched video.

[thetoof]

But how to load cpu closer to 90-100% ?

I have an idea to start 3 processes at once with different episodes :)Yup, that'd be the way to go. Or run different processes with different sections of the same episode (trim+rendering passes and splice them in the "to encode" script)

You can make a separate thread and post a sample of untouched video.
He could post it here since it's about help with animeivtc.

zocz, DGIndex-select a section-save and demux video-upload .m2v @ mediafire.com

[zocz]

thetoof thanks!
I upload it in the evening.

[zocz]

He could post it here since it's about help with animeivtc.

zocz, DGIndex-select a section-save and demux video-upload .m2v @ mediafire.com

Hi!
I uploaded files:
film1.m2v (http://www.mediafire.com/?hhjzhjmdyej) film1.d2v (http://www.mediafire.com/?neyqtnzd42k)
film2.m2v (http://www.mediafire.com/?o2jn124wbjm) film1.d2v (http://www.mediafire.com/?zxyodwjnngi)

I think, the movies are DHT, but I don't know.
Thanks for your help.