AnimeIVTC 2.X with bug fix and more

[real.finder]

hi

I have this edition of AnimeIVTC, which have many fix and Additions

Compared to the original version:-

1- chrfix really work now, even with mode 2

test successfully with http://forum.doom9.org/showthread.php?t=167194

sample http://forum.doom9.org/showthread.ph...56#post1615656

edit:- try not to use chrfix because it do not use IVTC and use deinterlacer and that will lose Quality unlike IVTC, use QTGMC_bob(0,0.5) like this

originali = last
ffapp = originali.QTGMC_bob(0,0.5).chroma_rep #or chroma_rep2 see dchr below
getparity(originali) ? ffapp.interlaced60or50 : ffapp.interlaced60or50(bff=true) #reinterlaced
ivtc here

2- add dchr bool, can help chrfix with some source

3- add extbob clip to use an External bobber with bbob/cbob=5, like this External bobber or External bobber with DeRainbow (feel free to change ASTDR settings to fit your source), EX:-

# start of the script, Including the source and trim and other
function MC_A_bob(clip input)
{
# stable and maintains details v1.11
TDeint_cilp=input.tdeint(1, tryWeave=true)
sup    = TDeint_cilp.HQdn3d.FFT3DFilter.MSuper(pel=2,sharp=1)
fv1    = sup.MAnalyse(isb=false,delta=1,DCT=2,Truemotion=false)
fv2    = sup.MAnalyse(isb=true,delta=1,DCT=2,Truemotion=true)
csaa   = input.QTGMC(lsbd=true, lsb=true, edimode="eedi3+nnedi3", SourceMatch=3, Lossless=2) # SourceMatch=3, Lossless=2 for less Artefacts in anime source, you can add tr0=1, tr1=1, tr2=1 too
momask1 = TDeint_cilp.MMask(fv1, kind = 1, ml=2)
momask2 = TDeint_cilp.MMask(fv2, kind = 1, ml=3)
momask = mt_average(momask1,momask2,u=3,v=3)
mt_merge(TDeint_cilp,csaa,momask,u=3,v=3)
}
AnimeIVTC(2, bbob=5, extbob=MC_A_bob, mode22=true)
# The rest of the filters

4- add bob4p to change QTGMC Preset

5- add ediext clip to use it with QTGMC (b/c bob 4)

6- TC file now will out in V2

7- add mode22 bool to use it in mode 2 for Spatial field-blend (field-blend in a part of the picture)

http://forum.doom9.org/showthread.php?t=169892

8- make "precision=1" use "nnedi3" only, for "Toriko Anime" (in op) or any 60i source, and also for speed

now it is:-

precision : Higher = slower, but more precise... Here, it's about quality vs speed and consonant in special cases.

0 : Fastest, tfm only
1 : Better (uses tfm + nnedi3 to deinterlace)
2 : Even better (tfm +Tdeint + nnedi3 for interpolation)
3 : Optimal (tfm +Tdeint + nned3i + motion mask) (default)

9- fix this http://forum.doom9.org/showthread.ph...37#post1621237

10- there are other by the thetoof and doom9 member in replies form original Thread

----------


some notes:-

1- try not to use aa, killcomb and iFade, Better alternative here

and for real Anti-aliasing use maa2 or maa2mod or maa2+ mod

2- in mode 2 and 3 (which work with bob), using daa (or some thing like it, like mcdaa3) and ifade almost useless in most cases, especially if the bob was good and stable, and the same thing when use chrfix in rest mode, Details in the following note

3- using chrfix will ignore precision (Field matching) and use cbob/bbob instead

4- you can use SEt avs 2.6 mt in mode 2 with most case in AnimeIVTC, test with this ver. http://www.dropbox.com/s/xhqggxamegi...th_20130309.7z

5- chrfix not fix Dot Crawl, the one who make it removal is QTGMC or TGMCmod, which is the default in cbob, and because mode 2 use bbob and the default is yadifmod not QTGMC, Dot Crawl will not fix by default

6- you can call AnimeIVTC many time in mode 1 and mode 3 (includes collect both) without chrfix, Preferably without anything, just mode number, like:-

trim(0,7899).animeivtc(1) +\
Trim(7900,8143).animeivtc(3) +\
Trim(8144,36413).animeivtc(1)

or

 a=Trim(0,7899).animeivtc(1)
 b=Trim(7900,8143).animeivtc(3)
 c=Trim(8144,36413).animeivtc(1)
 a+b+c

In order to keep the synchronization, trim should start with the beginning of the scene, and the end of trim ends with the scene end, Excluded from that the beginning and end of the episode, of course

and here a script to make a trims base on scenes changes


-------------


Small tip

if you have telecine source with Rainbow and\or Dot Crawl and TComb not enough for it you can use this

sourcefilter
TComb 
MC_A_bob #MC_A_bob with DeRainbow edition
srescr = last
srestore(frate=23.976, mode=-4, thresh=44, cache=10, dclip=srescr.crop(10,0,-10,0).reduceflicker(strength=1)) #or srestore(frate=23.976, thresh=44, cache=10, dclip=srescr.crop(10,0,-10,0)) if you don't use MC_A_bob with DeRainbow

without animeivtc and without daa\daa3

edit:- see this http://forum.doom9.org/showthread.php?t=172221 it better

now there are ASTDRmc ver. and it is advisable to use it after the restore to 24 (23.976) fps clip, and it should be better than ASTDR with bob clip in avoid chroma bleeding in moving scenes

---

* if you have a telecine source with a light Rainbow, even if you did not notes it, it maybe destroy IVTC Filter accuracy in some frames, so you must remove that Rainbow before IVTC, use TComb(1) can help with that

and some time the scan order wrong, you can check it by SeparateFields() and see the moving scenes, if it wrong correct it by put AssumeBFF() or AssumeTFF() after source calling line

more notes in page 3 1st post http://forum.doom9.org/showthread.ph...35#post1720135

*try not to use anything use bob because it do not use IVTC and use deinterlacer and that will lose Quality unlike IVTC, even for Field blended you can do this http://forum.doom9.org/showpost.php?...9&postcount=49

----------


download links:- link

and in attachments too this old version

newer versions and notes in 2nd post

[real.finder]

if you Looking for AnimeIVTC helper (AnimeIVTC_helper) it's here

2014-3-18:- add palf bool for no decimate in mode 1 and make fps 24 instead of 25 for mode 2 in PAL source, for sources that transferred from film directly to PAL, download pastebin

2014-3-19:- add tcfv1 bool to made TC file v1 and make it true by default, download pastebin

2014-3-23:- add nvfr bool for VFR noisy sources, small change in relationship between ifade and killcomb, download pastebin

2014-3-28:- some edit in mode22, download pastebin

2014-3-31:- add real30p bool to not use daa with 30p, and some fix for PAL cases and other things, download pastebin

2015-3-29:- add TFM (mode, cthresh, blockx, blocky, MI) and some fix, download pastebin, and in Attachments below

2015-4-15:- add extbobf string to use external bob function, old external bob clip (extbob) still exist, and if you use both the old extbob will win, download pastebin, and in Attachments below

2015-5-02:- make precision=0 use pp=0 in tfm for speed and quality too, add precision=4 same as precision=3 but no blend, add precision=5 to use yadifmod (usefull for very old anime), add precision=6 to use extbobf\extbob with ConvertFPS to back to original fps, add precision=7 to use extbobf\extbob with SelectEven to back to original fps (no blend), and precision=8 for already half bob fps, download pastebin

2015-5-07:- add autoAssuf to auto set field to match from in tfm to help with noisy old telecined source, add tfm pp and metric, download pastebin, and in Attachments below

2015-10-10:- tweak autoAssuf in mode=1 with another method, others modifications; add tfm micmatching for it important to set it to 0 (zero) in many dvd 20XX anime (maybe other source need that too); add ediandnn to chose between nnedi3 if set to 1 or eedi3 if set to 2 or eedi3+nnedi3 if set to 3, others modifications; add yadifrepwithtdeint to use yadifmod with less Artefacts; add o3025cfr to out 25 in PAL and 30 in NTSC, and it useful to avoid blended with omode=1; palf and o3025cfr more tweak, now palf work in pal only and in mode 1 and 3 will out 24; fix nnedi3_rpow2 and make credconv accept external clip for credits; workaround for RgTools Repair in yadifrepwithtdeint; add bobpresmooth for more stability. download pastebin

2015-10-28:-
# add bbob4_params and cbob4_params to use more settings for QTGMC, add precision=9 to use bbob (maybe you need it with bbob=4 to use QTGMC with TFM)
# add ediandnn=4 to use QTGMC in yadifmod in yadifrepwithtdeint , others
# make precision=6 TDeint + motion mask , add b/c bob 6 TDeint + motion mask
# chrfix now work with YUY2 in avs 2.6
# yadifrepwithtdeint now repwithtdeint and work with bbob 3 and 4 and precision 5
# add crepwithtdeint for cbob 3 and 4
# tweak autoAssuf
# add tdeint sharp bool and type int to set interpolation for c/repwithtdeint and b/cbob precision 6. download pastebin, and in Attachments below

2016-10-14:-
# update chroma_rep/chroma_rep2 to http://forum.doom9.org/showthread.ph...ep#post1268074 (this not work without some edit I did in 2014), chrfix now will accept float as thr, add mChroma/chroma tfm
# ediext clip now can be used instead of ediandnn, add prefiltered_i to use it instead of bobpresmooth. download pastebin

v2.20 2017-02-09:-
# add yadifmod2 and other things for yadifmod, make the cache more for mode22 and chrfix
# add tmm2. download pastebin, and in Attachments below and mediafire

v2.21 2018-02-28:-
# added vidDetect parameter for VFR. download pastebin, and in Attachments below

v2.22 2018-04-17:-
# added mode 5 same as 3 but use nnrep method to ediandnn, mode 6 too, but it's like 4, and now you can use negative value for use nnedi3 with Padding to avoids edges distortion on top and bottom. download pastebin, and in Attachments below

last versions includes sanimebob will be here https://github.com/realfinder/AVS-St...AnimeIVTC.avsi

Manual/Guide http://avisynth.nl/index.php/AnimeIVTC (may have outdated info)

[Overdrive80]

Thanks for you effort and contribution. ^^

[real.finder]

Originally Posted by Overdrive80

Thanks for you effort and contribution. ^^

You are welcome

-----


another new release pending approval in attachments

[Guest]

Originally Posted by real.finder

another new release pending approval in attachments

Approved. Thank you for your contributions to the forum. Feel free to PM me for immediate attachment approval.

[osgZach]

Cool, something new to play with.. certainly interested to see if the chroma fixes will help with a particular video.

Any chance on a rewritten/updated manual to reflect these changes?

[real.finder]

thanks neuron2, I will

Originally Posted by osgZach

Any chance on a rewritten/updated manual to reflect these changes?

Maybe I updated it later, Until then, any volunteer to update it are welcome, as my English is not good

[osgZach]

I noticed something strange with the script.

Assuming you haven't messed around with the VFR modes, I pretty much ran a test with a script I had used with the old version of the script, using mode 4 with a 1-st (analyze) and 2nd (render) pass and I got completely different timecodes and footage identity. As far as I am aware I didn't really tweak much, except maybe did more or less filtering on certain things. i.e Original I used AA=4 (sharpAAMCMod) but don't recall if I ran that separate or not. This version I did no AA, but I did do some degraining and used a higher level of killcomb I believe..

But generally I kept the same mode settings, using the same source D2V and DVD rip files. Ran 1st passes with no filtering for speed, and used filtering on 2nd passes. Your script was a lot more CPU hungry but that could have been to different settings I specified like degraining and such.

Under the old Script I got a video output with around 24.5xx something FPS, Timecode default FPS of 30, with 23.96 and 17.xx something hinted sections. The split was around 89.x% film 10.x% video

Under this one I got a video that was closer to 23.976 FPS with a default FPS of 30 and hinted sections at 23.967 and a lot more 17fps sections. It was pretty much an odd/even pattern (23/17/23/17) as far as the frame range entries went. The video produced by this script was noticeably jerky on shots that were otherwise smooth with the original version of AnimeIVTC as well. The entire clip was identified at 100% Film.

I have to be honest that I didn't go into the whole affair intending to take notes about things, so if there is some specific information you would like me to gather and report on, or any specific script settings to enable/disable please let me know and I will be happy to do re-encodes using both versions of the script to see what I can come up with.

[real.finder]

Originally Posted by osgZach

I noticed something strange with the script.

Assuming you haven't messed around with the VFR modes, I pretty much ran a test with a script I had used with the old version of the script, using mode 4 with a 1-st (analyze) and 2nd (render) pass and I got completely different timecodes and footage identity. As far as I am aware I didn't really tweak much, except maybe did more or less filtering on certain things. i.e Original I used AA=4 (sharpAAMCMod) but don't recall if I ran that separate or not. This version I did no AA, but I did do some degraining and used a higher level of killcomb I believe..

But generally I kept the same mode settings, using the same source D2V and DVD rip files. Ran 1st passes with no filtering for speed, and used filtering on 2nd passes. Your script was a lot more CPU hungry but that could have been to different settings I specified like degraining and such.

Under the old Script I got a video output with around 24.5xx something FPS, Timecode default FPS of 30, with 23.96 and 17.xx something hinted sections. The split was around 89.x% film 10.x% video

Under this one I got a video that was closer to 23.976 FPS with a default FPS of 30 and hinted sections at 23.967 and a lot more 17fps sections. It was pretty much an odd/even pattern (23/17/23/17) as far as the frame range entries went. The video produced by this script was noticeably jerky on shots that were otherwise smooth with the original version of AnimeIVTC as well. The entire clip was identified at 100% Film.

I have to be honest that I didn't go into the whole affair intending to take notes about things, so if there is some specific information you would like me to gather and report on, or any specific script settings to enable/disable please let me know and I will be happy to do re-encodes using both versions of the script to see what I can come up with.

yes, that in "10- there are other by the thetoof and doom9 member in replies form original Thread"

There is a change in the VFR

You can use winmerge To see the changes between the original script and this

the change is

finalvfr = 	omode==2 && pass==1 ? tovfrdec1.TDecimate(4, output="stats.txt",denoise=true,vidThresh=0.8) : 
\		omode==2 && pass==2 ? tovfrdec2.TDecimate(5,hybrid=2,tfmin=creds ? "" : "tfm.txt",input="stats.txt",mkvout="timecodes.txt",tcfv1=tcfv1,batch=true,vidThresh=0.2) : nop()

from TDecimate - READ ME.txt

vidThresh -

This setting is used for detecting video sections (30p) based off frame metrics
when hybrid > 0. If all frames in a cycle have metrics above this threshold then
the cycle is declared video metrics wise. This setting is similar to dupThresh,
but should be set slightly higher if your vidDetect setting is set to 1 and not 3.
If you know that your source has a lot of video sequences then set this lower vs if
you know your source is pure film then you can set this really high to prevent any
possible misdetections. This value is a % of maximum change for a block defined by
the blockx and blocky values.... so 1.1 means 1.1% of maximal possible change.

When in mode 7, vidThresh is used to set a threshold above which frames are considered
to absolutely be non-duplicates. Use display=true to see the unnormalized threshold
values and the frame difference values.

Default: 1.1 (if chroma = true) (float) -- modes 1, 3, 5 && hybrid > 0
1.4 (if chroma = false)

3.5 (if chroma = true) (float) -- mode 7
4.0 (if chroma = false)

denoise -

Sets whether or not to denoise frames prior to doing difference metric calculations.
This can greatly improve metrics for noisy sources (i.e. stabilize duplicate frame
metrics and make it easier to tell the difference between duplicates and non-duplicates).
It also works very well for sources with lots of dot-crawl because the denoising
effectively cancels all dot-crawl, whereas the dot-crawl would usually inflate difference
values of duplicates. Using denoising will slow things down somewhat, but it is MMX
optimized and pretty fast.

true - use denoising
false - don't

Default: false (bool)

so, I find this changes (Which I found it in Earlier ver in replies) good, and use it in one piece and the result was good.

But, if you see it bad, I can Make it optional and the default like the original script, Can you upload a sample from the clip you are working on?

[real.finder]

tweaked and added nvfr bool to use the VFR changes mentioned above

nvfr is false by default, so the result in mode 4 (vfr) will be like the original script by default

----


test script with 17/23/29 fps

a=ColorBars(pixel_type="yv12").trim(1, 1).KillAudio()
b=BlankClip(pixel_type="yv12", fps=30000./1001).trim(1, 1).KillAudio()

a+b+b+a+b + a+b+b+a+a + b+a+b+a+a + b+a+b+a+b + a+b+a+b+b

AddGrainC.AddGrainC.AddGrainC

AnimeIVTC(4, omode=2, pass=2 ,nvfr=true)

[osgZach]

Working on running tests now, but had a question incase anyone knows the answer..

When you demux an m2v from a VOB with DGindex, is the M2V supposed to have pulldown flags, etc if there were originally present? (Is it the VOB or the M2V which holds such flags?)

[Guest]

Originally Posted by osgZach

When you demux an m2v from a VOB with DGindex, is the M2V supposed to have pulldown flags, etc if there were originally present? (Is it the VOB or the M2V which holds such flags?)

The elementary stream contains the pulldown flags, so if you demux they will still be present.

[osgZach]

Hmm my first round of tests produced seemingly similar results. At least I think within natural margins of difference. 0.020FPS difference on output AVIs. There are some Film/Video % differences (0.xx%) and there are some timecode differences but not many. The instance where a ship moving across the screen was jerky is now smooth even under the new version encode..

I will have to test with more filter options enabled I think, and see if that changes anything. If there was an error perhaps it was fixed in the newer most recent version of your script. I may have to go back and test the older version I initially used. Will continue to report back my findings.

edit: It's definitely filters causing some kind of change. The timecodes output for the clip just changed to 100% Video, and all I did was add back the filters I was using

AnimeIVTC(mode=4, pass=2, omode=2, nvfr=true, chrfix=true, tcfv1=true, killcomb=3, degrain=3, aa=0, rendering=false)

The culprit appears to be chrfix=true
Once I took it out, I got a much more proper timecodes file, in line with my earlier non-filtered tests. I don't recall reading anything about chrfix messing up VFR modes, am I mistaken?

[real.finder]

Originally Posted by osgZach

Hmm my first round of tests produced seemingly similar results. At least I think within natural margins of difference. 0.020FPS difference on output AVIs. There are some Film/Video % differences (0.xx%) and there are some timecode differences but not many. The instance where a ship moving across the screen was jerky is now smooth even under the new version encode..

I will have to test with more filter options enabled I think, and see if that changes anything. If there was an error perhaps it was fixed in the newer most recent version of your script. I may have to go back and test the older version I initially used. Will continue to report back my findings.

edit: It's definitely filters causing some kind of change. The timecodes output for the clip just changed to 100% Video, and all I did was add back the filters I was using

AnimeIVTC(mode=4, pass=2, omode=2, nvfr=true, chrfix=true, tcfv1=true, killcomb=3, degrain=3, aa=0, rendering=false)

The culprit appears to be chrfix=true
Once I took it out, I got a much more proper timecodes file, in line with my earlier non-filtered tests. I don't recall reading anything about chrfix messing up VFR modes, am I mistaken?

using chrfix will have an effect, see note 3 in this Thread

with nvfr the result will be different from the original script, indeed

and by the way, you don't have to use killcomb if you use mcdaa3

[osgZach]

You have 2 sets of notes - I assume you mean note 3 from the second set about it bobbing the clip?

So, if I wanted to use chrfix, is there any way to get around this issue?

[real.finder]

Originally Posted by osgZach

You have 2 sets of notes - I assume you mean note 3 from the second set about it bobbing the clip?

So, if I wanted to use chrfix, is there any way to get around this issue?

yes, the first one is for the changes between this ver. and the original one

As I say, using chrfix in whether the original script or this will cancel precision (Field matching) and use cbob/bbob instead

chrfix = chroma_rep

try nvfr=false in animeivtc may help you

and are you really need chrfix? it for state like this http://img27.imageshack.us/img27/6109/chromareptest0io1.jpg

http://uppix.net/e/8/2/9f9661a24ae29a87ea64ec3012d2d.png

http://imageshack.us/a/img194/7543/vts011092198.png

http://img203.imageshack.us/img203/1695/afterivtc.png

[osgZach]

I believe I may. Out of all those pictures, the final one (afterivtc.png) looks closest to what I believe I am also seeing in my source clip.

It has rainbowing issues I can't seem to fix regardless, but I was trying to clean up issues like colors on black line edges, etc. Like the clip shows with the colors around the police officers back.

It is not a pristine DVD transfer by any means ( looks like a poor VHS copy), so it's not the end of the world, if I can't do anything to clean that stuff up.

[real.finder]

Originally Posted by osgZach

I believe I may. Out of all those pictures, the final one (afterivtc.png) looks closest to what I believe I am also seeing in my source clip.

It has rainbowing issues I can't seem to fix regardless, but I was trying to clean up issues like colors on black line edges, etc. Like the clip shows with the colors around the police officers back.

It is not a pristine DVD transfer by any means ( looks like a poor VHS copy), so it's not the end of the world, if I can't do anything to clean that stuff up.

Is it possible to upload a sample to see and enjoy a little bit with this source?

[osgZach]

I'll see if I can find something suitable and cut a sample

[osgZach]

Ok Here is my sample.

I didn't really know the best way to cut it, so I opened my demuxed m2v in DGindex and ran a selection then saved out a D2V with the smaller cut Mpeg.

This section appears to be entirely film (DGindex says) but I the entire movie has 30p sections in it, so its not a simple IVTC. I would not approach it from that solution.

I believe this short clip will show the color issue around character lines, I thought looked similar to the clip mentioned above with the police officer.

Gall Force: Eternal Story, sample clip
(25mb)