Help with mixed cadence in old anime DV

[RItas]

I’m completely new to video encoding and I’m trying to back up an old anime DVD. I’ve been using StaxRip, mostly fumbling my way through it with the help of AI, so please bear with me if I get some terms wrong.
I have two separate issues with different title sets from an old anime DVD.


VTS_01 (main animation)

After separating fields with SeparateFields(), I noticed this pattern on the main animation content: every 5 fields, the first 2 fields are identical, and the next 3 fields are identical (AA BBB). This looks like a standard 3:2 pulldown from 24p.

However, the episode also has an animated logo and some staff credits that fade in and out. In those parts, every single field is different – true 30i. The tricky part is when the logo/credits overlap with the animation content; both patterns appear on screen at the same time. The background animation still follows the 2:3 pattern, while the overlays have unique fields.

I’m not sure how to approach this kind of mixed content. What would be the recommended way to handle it and get clean progressive frames? I’d appreciate any advice.


VTS_02 (interview)

After SeparateFields():


The live-action interview part: every field is different.



The animated logo pop-up part (only at the beginning): every two adjacent fields are identical.


I processed this VTS_02 with QTGMC in StaxRip (AviSynth). After processing, in the logo pop-up part, the edges of the logo wobble/shift between consecutive frames.

Is there a way to fix this?

[hello_hello]

You probably should upload some small samples somewhere and post a link.

[RItas]

You probably should upload some small samples somewhere and post a link.

Thank you for the reply, and my apologies — I wasn't aware that providing samples was expected. I've already cut a few VOB clips showing the problematic sections. Here they are: MEGA (https://mega.nz/folder/mahRQTRA#w8ZoaEtRnM2pLATztjovPQ)

[Emulgator]

VTS_01: Bob it up to 59,94p using QTGMC or QTGMCp
VTS_02: Bob it up to 59,94p using QTGMC or QTGMCp

Like:

AssumeTFF()#.SeparateFields()
#QTGMC(EdiMode="BWDIF+NNEDI3", tr2=3, sharpness=1.0, Lossless=2, SourceMatch=3, Sbb=1, ShowSettings=false, Denoiser="KNLMeansCL", NoiseTR=2)
propSet("_FieldBased",2) #QTGMCp braucht das unbedingt ! #Oder propClearAll(last)
QTGMCp(InputType=0,EdiMode="BWDIF", tr2=3, Rep1=2, RepChroma=true, ChromaMotion=false, DenoiseMC=true, sharpness=2.0, lossless=0, NoiseProcess=1, GrainRestore=0.0, NoiseRestore=0.5, ChromaNoise=true, show=false)

I just left my defaults, tailoring any parameters is up to you.
Well and healing the blocking bubbles from VTS_01: Oyster.
For such an animation of swarming elements the MPEG2 encoder can not find any sane motion vectors.
This would have needed much more bits, even at full DVD bitrate this encode came out starved.
VTS_02: The camera footage needs Dehaloing.

[hello_hello]

For "VTS_01 sample.VOB", assuming the "interlaced over telecine" problem is only at the beginning and the bulk of it's telecined, I'd consider keeping it simple and field match and deinterlace with TFM, then blend decimate the sections with an interlaced overlay to 23.976fps. It should ensure the motion remains smooth.

Mpeg2Source("D:\VTS_01 sample.d2v")
TFM(Metric=1, pp=5)
TDecimate(Mode=1, Hybrid=1)
CropResize(0,0, 8,0,-8,-0, InDAR=15.0/11.0, ResizeWO=true) # optional, of course

VTS_01 sample.mkv (https://files.videohelp.com/u/210984/VTS_01%20sample.mkv)

"VTS_02 sample.VOB" seems to be a job for QTGMC.

QTGMC()
# SelectEven() for 30fps rather than 60fps
CropResize(640,480, 10,0,-8,-0, InDAR=15.0/11.0) # optional, of course
FineDehalo()
FineDehalo()

VTS_02 sample.mkv (https://files.videohelp.com/u/210984/VTS_02%20sample.mkv)

[RItas]

VTS_01: Bob it up to 59,94p using QTGMC or QTGMCp
VTS_02: Bob it up to 59,94p using QTGMC or QTGMCp

Like:

AssumeTFF()#.SeparateFields()
#QTGMC(EdiMode="BWDIF+NNEDI3", tr2=3, sharpness=1.0, Lossless=2, SourceMatch=3, Sbb=1, ShowSettings=false, Denoiser="KNLMeansCL", NoiseTR=2)
propSet("_FieldBased",2) #QTGMCp braucht das unbedingt ! #Oder propClearAll(last)
QTGMCp(InputType=0,EdiMode="BWDIF", tr2=3, Rep1=2, RepChroma=true, ChromaMotion=false, DenoiseMC=true, sharpness=2.0, lossless=0, NoiseProcess=1, GrainRestore=0.0, NoiseRestore=0.5, ChromaNoise=true, show=false)

I just left my defaults, tailoring any parameters is up to you.
Well and healing the blocking bubbles from VTS_01: Oyster.
For such an animation of swarming elements the MPEG2 encoder can not find any sane motion vectors.
This would have needed much more bits, even at full DVD bitrate this encode came out starved.
VTS_02: The camera footage needs Dehaloing.

For "VTS_01 sample.VOB", assuming the "interlaced over telecine" problem is only at the beginning and the bulk of it's telecined, I'd consider keeping it simple and field match and deinterlace with TFM, then blend decimate the sections with an interlaced overlay to 23.976fps. It should ensure the motion remains smooth.

Mpeg2Source("D:\VTS_01 sample.d2v")
TFM(Metric=1, pp=5)
TDecimate(Mode=1, Hybrid=1)
CropResize(0,0, 8,0,-8,-0, InDAR=15.0/11.0, ResizeWO=true) # optional, of course

VTS_01 sample.mkv (https://files.videohelp.com/u/210984/VTS_01%20sample.mkv)

"VTS_02 sample.VOB" seems to be a job for QTGMC.

QTGMC()
# SelectEven() for 30fps rather than 60fps
CropResize(640,480, 10,0,-8,-0, InDAR=15.0/11.0) # optional, of course
FineDehalo()
FineDehalo()

VTS_02 sample.mkv (https://files.videohelp.com/u/210984/VTS_02%20sample.mkv)

Thank you both! Just to clarify: the mixed sections are only at the very beginning for both VTS_01 and VTS_02.

VTS_01: after that, all standard 3:2 pulldown (three identical fields, then two).

VTS_02: after the logo, pure 30i.

In the VTS_01 sample encode shared by hello_hello, I still see combing around the 8-second mark where the light particles gather into the credits. Is there a way to fix that?

[hello_hello]

Stepping through the frames, I could only see one frame with actual combing. Lowering the default CThresh value seems to do the trick (looking at my new sample again there may be a few combed dots where they fade in if you want to try an even lower CThresh value).

I used pp=5 for TFM() as it uses blend de-interlacing. The default of pp=6 created noticeable aliasing in another section (as did QTGMC with it's default values). PP=5 does blur the moving dots in the section you're referring to though, so it might be better to do a little manual trimming and appending to produce the best possible output. You could create 3 or maybe 4 clips that do things differently, then just pick the clip to use for certain frame ranges (probably best to switch clips on a scene change). Something like:

# Blend de-interlacing and blend decimation to 23.976fps
ClipA = TFM(Metric=1, PP=5, CThresh=5).TDecimate(Mode=1, Hybrid=1)

# Cubic interpolation deinterlacing and blend decimation to 23.976
ClipB = TFM(Metric=1, PP=6, CThresh=6).TDecimate(Mode=1, Hybrid=1)

# Blend de-interlacing and 1 frame in 5 decimation for the purely telecined sections
ClipC = TFM(Metric=1, PP=5, CThresh=6).TDecimate(Mode=1)

# Cubic interpolation deinterlacing and 1 frame in 5 decimation for the purely telecined sections
ClipD = TFM(Metric=1, PP=6, CThresh=6).TDecimate(Mode=1)

ClipB.Trim(0, 478) + \
ClipA.Trim(479, 622) + \
ClipB.Trim(623, 747) + \
ClipC.Trim(748, 0)

CropResize(0,0, 8,0,-8,-0, InDAR=15.0/11.0, ResizeWO=true)

VTS_01 sample B.mkv (https://files.videohelp.com/u/210984/VTS_01%20sample%20B.mkv)

If it's only the intro you have to check manually it shouldn't be too hard (although I don't use Staxrip so I don't know how easily it lets you modify or create your own Avisynth scripts).
Click on the thumbnails for the full screenshots.

PP=5, Frame 209
https://i.ibb.co/jZfjDyY2/PP5.png (https://ibb.co/jZfjDyY2)

PP=6, Frame 209
https://i.ibb.co/mCsdFHfh/PP6.png (https://ibb.co/mCsdFHfh)

Here, I think the blend deinterlacing is the lesser of 2 evils, given the dots are moving pretty quickly anyway.

PP=5, Frame 514
https://i.ibb.co/LzQ3y8yk/PP5B.png (https://ibb.co/LzQ3y8yk)

PP=6, Frame 514
https://i.ibb.co/vvwYCXvH/PP6B.png (https://ibb.co/vvwYCXvH)

[RItas]

Stepping through the frames, I could only see one frame with actual combing. Lowering the default CThresh value seems to do the trick (looking at my new sample again there may be a few combed dots where they fade in if you want to try an even lower CThresh value).

I used pp=5 for TFM() as it uses blend de-interlacing. The default of pp=6 created noticeable aliasing in another section (as did QTGMC with it's default values). PP=5 does blur the moving dots in the section you're referring to though, so it might be better to do a little manual trimming and appending to produce the best possible output. You could create 3 or maybe 4 clips that do things differently, then just pick the clip to use for certain frame ranges (probably best to switch clips on a scene change). Something like:

# Blend de-interlacing and blend decimation to 23.976fps
ClipA = TFM(Metric=1, PP=5, CThresh=5).TDecimate(Mode=1, Hybrid=1)

# Cubic interpolation deinterlacing and blend decimation to 23.976
ClipB = TFM(Metric=1, PP=6, CThresh=6).TDecimate(Mode=1, Hybrid=1)

# Blend de-interlacing and 1 frame in 5 decimation for the purely telecined sections
ClipC = TFM(Metric=1, PP=5, CThresh=6).TDecimate(Mode=1)

# Cubic interpolation deinterlacing and 1 frame in 5 decimation for the purely telecined sections
ClipD = TFM(Metric=1, PP=6, CThresh=6).TDecimate(Mode=1)

ClipB.Trim(0, 478) + \
ClipA.Trim(479, 622) + \
ClipB.Trim(623, 747) + \
ClipC.Trim(748, 0)

CropResize(0,0, 8,0,-8,-0, InDAR=15.0/11.0, ResizeWO=true)

VTS_01 sample B.mkv (https://files.videohelp.com/u/210984/VTS_01%20sample%20B.mkv)

If it's only the intro you have to check manually it shouldn't be too hard (although I don't use Staxrip so I don't know how easily it lets you modify or create your own Avisynth scripts).
Click on the thumbnails for the full screenshots.

PP=5, Frame 209
https://i.ibb.co/jZfjDyY2/PP5.png (https://ibb.co/jZfjDyY2)

PP=6, Frame 209
https://i.ibb.co/mCsdFHfh/PP6.png (https://ibb.co/mCsdFHfh)

Here, I think the blend deinterlacing is the lesser of 2 evils, given the dots are moving pretty quickly anyway.

PP=5, Frame 514
https://i.ibb.co/LzQ3y8yk/PP5B.png (https://ibb.co/LzQ3y8yk)

PP=6, Frame 514
https://i.ibb.co/vvwYCXvH/PP6B.png (https://ibb.co/vvwYCXvH)

Thank you for the detailed explanation! I agree that keeping the animation content clear is more important than making those light particles sharper. I'll try encoding the whole video with pp=5 and pp=6 separately, compare the results, and then decide which way to go.

[hello_hello]

In theory, if TFM can field match in the telecined sections and restore the original frames, pp=5 or pp=6 shouldn't matter because there'd be nothing for it to de-interlace if the field matching results in clean progressive frames as there'd be no combing. It only de-interlaces where it detects combing, where field matching doesn't result in a clean frame for some reason or where the content is interlaced rather than telecined.