Wise choice to deinterlace that DVD material?

[Weef]

Now looking at the filters options of TIVTC and the mentioned situation that the animations are done in different amounts of fps.
Should I use a hybrid option?

[poisondeathray]

Now looking at the filters options of TIVTC and the mentioned situation that the animations are done in different amounts of fps.
Should I use a hybrid option?

The samples from this source you've posted don't look like they require any special treatment, just regular IVTC

[poisondeathray]

It sounds like you're fairly new to avisynth , To speed things up, I would use DGIndex and TIVTC

The first part of your script will look like

MPEG2Source("video.d2v")
AssumeTFF
TFM()
TDecimate()

So that will return the original progressive film frames, 23.976p . You add whatever filters afterwards




Links to some learning resources

http://avisynth.org/mediawiki/Main_Page#New_to_AviSynth_-_start_here
http://neuron2.net/faq.html#analysis

[manono]

The first part of your script will look like

MPEG2Source("video.d2v")
AssumeTFF
TFM()
TDecimate()

For anime specifically it's often helpful to use TDecimate(Mode=1). From the TDecimate doc:

Mode 1 = Exactly like mode 0, except instead of decimating the M most similar
frames, frames are decimated from the longest remaining strings of
duplicates. The duplicate detection uses the dupThresh parameter. This
mode is the correct type of decimation for anime and other sources where
frames are repeated 2, 3 or 4 times in a row

Newcomers sometimes dive into AnimeIVTC and YATTA or use hybrid options when there's no need to and needlessly complicate things and sometimes wind up with poorer results than had they used a simple TFM.TDecimate.

[Weef]

Since I will repackage it as .mkv and softsub it with an .ass file, is there anything else I should be taking into account?

[Weef]

Another question popped into my mind.
Since I at first shouldn't export from Virtual Dub and or anything else but directly via AviSynth.
Does AviSynth have a better resizer than Virtual Dub?
And why do we take 856x480?
Wouldn't the multiple of 16 be 848?
Or have you already taken into account that black bar....that only appears in some episodes and therefore added some pixels to cute them out of the frame later?

[poisondeathray]

Another question popped into my mind.
Since I at first shouldn't export from Virtual Dub and or anything else but directly via AviSynth.
Does AviSynth have a better resizer than Virtual Dub?
And why do we take 856x480?
Wouldn't the multiple of 16 be 848?
Or have you already taken into account that black bar....that only appears in some episodes and therefore added some pixels to cute them out of the frame later?

It's better to do everything in avisynth (same colorspace) if you can, because most vdub filters require you to convert to RGB

Avisynth has better resizers, more control, and more options



It's up to you - mod16 will give you slightly more efficient compression . Some filters have certain mod restrictions

In general, in terms of compression efficiency , and compatibility with filters and some decoders : mod16 > mod8 > mod4 > mod2

16/9 = 1.7777

854x480 is the closest in AR with mod2 compliance 854/480 = 1.7791
848x480 is mod16 compliant but AR is off more = 1.7666
856x480 is mod8 compliant but AR is off more = 1.7833

Most people won't notice the <1% difference in AR values . (But a bit off topic - if you upload 848 to youtube it will be slightly pillarboxed, 856 will be letterboxed, 854 will fit perfect)

[Weef]

Wow, and in mod2...how many filters can I still use? ^^"

[poisondeathray]

Wow, and in mod2...how many filters can I still use? ^^"

I don't know, but some of them require mod4 or better .

The AR difference is neglible when viewed by itself . You were going to go with oval planets at first so you probably really don't care about the slight AR differences .

[Weef]

That's so true xD

[Weef]

Ok, one last question to this topic.
As I read through the pages I came across the Spline resizers.
Now of course, not mentioning some extra work, I'm asking myself, if I shouldn't put some extra work in the project to make it worthwhile and do an upscale of this anime (720p).
Using the the bitrate calculator of videohelp, I now assume, I would at least need 4300kbps in video bitrate to keep it decent (0,195 bits per pixel).
Is that a realistic expectation that it could still look good?

[manono]

the B/P*F figure means next to nothing. Do a single-pass CRF encode for whatever quality you want.

...but the export of the intro and credits has to go the way from Adobe products.... because I did it there and don't have a better program for that purpose.

You have AviSynth.

[jagabo]

the B/P*F figure means next to nothing.

Agreed. See the videos in this post and decide how meaningful that value is:
https://forum.videohelp.com/threads/295672-A-problem-for-video-experts?p=1811057&viewfu...=1#post1811057

Do a single-pass CRF encode for whatever quality you want.

And agreed.

[Weef]

So I did some try-and-error tests with MeGUI... just really basic stuff to get to know it better for the next works.
And I actually did some upscaling tests even the export options of MeGUI still confuse me bit and since I want to have it finished during the next month, I once again have to ask you guys ^^".

These are the settings I finally used for the last test

ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L3.1
Format settings, CABAC : Yes
Format settings, ReFrames : 6 frames
Format settings, GOP : M=4, N=16
Codec ID : avc1
Codec ID/Info : Advanced Video Coding
Duration : 18mn 15s
Bit rate mode : Variable
Bit rate : 1 899 Kbps
Maximum bit rate : 16.8 Mbps
Width : 1 108 pixels
Height : 608 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 23.976 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.118
Stream size : 249 MiB (100%)
Writing library : x264 core 130 r2273 b3065e6
Encoding settings : cabac=1 / ref=6 / deblock=1:1:1 / analyse=0x3:0x133 / me=umh / subme=10 / psy=1 / psy_rd=0.40:0.00 / mixed_ref=1 / me_range=24 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-2 / threads=12 / lookahead_threads=3 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=10 / b_pyramid=2 / b_adapt=2 / b_bias=0 / direct=3 / weightb=1 / open_gop=0 / weightp=2 / keyint=240 / keyint_min=23 / scenecut=40 / intra_refresh=0 / rc_lookahead=60 / rc=2pass / mbtree=1 / bitrate=1899 / ratetol=1.0 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / cplxblur=20.0 / qblur=0.5 / vbv_maxrate=17500 / vbv_bufsize=17500 / nal_hrd=vbr / ip_ratio=1.40 / aq=1:0.60

I'm posting these in conjunction with my coming questions so we can keep this short, if should have understand anything wrong.

So as you see I really did some upscaling but not to 720p material.
The reason this time I kept it at 1108x608 (or 1104x608 what the final result will probably be unless x264 has no problem with 1108x608 as it now seems (I thaught most codecs would have problems with anything under 16pixel steps sizes o.o )), is that with the tested CBR of 1900 I seem to get decent results in the scenes with a bit more than usual movement for that anime and the lines don't tend to get blocky. Also I don't want to end up with more of 350MB per 24Min episode.

With variable bitrate that was reduced to about 780 by x264 itself (since the update the interface seems to have changed) I ended up with unhomogenous impressions of color (blocky) in places where it should just stay all the same.
And in more rapid movements there were more blocks which blurred the image.

Now to my targeted questions.

1. Is there any way to frameserve my changes to the format from MeGUI directly to Premiere Pro OR to choose any more direct export from MeGUI? Should I really encode with the only two given options in MeGUI Xvid or x264 for export?

2. At this resolution I get a quick rendering (much faster than what I expected with Spline) but in some really rare moment in certain angle and color combinations I still see some....lets's not say jaggy.... but incosistent lines (neck and nose).
Should I try to add an antialising filter?

PS I did that render test on a 18min long sequence
and I'm still confused if I now really did a constant bit rendering or variable since it says yes but I targeted 1900...

[poisondeathray]

1) There is an avs import plugin but it is not reliable for CS5/6 because you need avisynth 64bit (which isn't reliable, very buggy) . You don't have to use megui, you can use the same script with vdub (video=>fast recompress to preven colorspace conversion) , or any other GUI front end that accepts avs scripts, and use a lossless intermediate e.g. ut video codec , for import into premiere. But why are you importing into premiere ???

2) AA should improve that but will always damage fine details to some extent . There are literally dozens you can choose from . AA should be applied before your upscale if you choose to use it - you have to play with them and the settings to determine which one works better for this source without damaging too much.

Instead I would try to determine what's causing it - e.g. is it really in the source, or maybe a bad IVTC (could you have used better settings ? maybe adjust the combing threshold ?) . Upscaling can cause jaggies too. A simple spline resize will cause jaggies but EDI based resizing (e.g nnedi3_rpow2) will not at the expense of fine details. Or maybe another option is selective filtering if it only affects certain sections - this way you don't damage the entire thing unnecessarily

Was this sequence uploaded in one of the samples ?

[Weef]

To answer

1. Does it work with CS4? I'm still using CS4.
Good to know!
I have to import it there first to add my credits and that intro. I could of course join them at the end....but somewhere and somewhen I have to join them with tht new rendered intro and credits.

2.Uhh, really before upscaling.... can I really finetune that far for those few specific sections :/
Where can I manipulate the TIVTC filter in MeGUI o.o??

And yes, can you give me link to an EDI dll plugin for MeGUI.
My MeGUI doesn't seem to open the filter when I call them...

[poisondeathray]

To answer

1. Does it work with CS4? I'm still using CS4.
Good to know!
I have to import it there first to add my credits and that intro. I could of course join them at the end....but somewhere and somewhen I have to join them with tht new rendered intro and credits.

yes, the old version works with CS4 - because CS4 is 32bit
http://videoeditorskit.sourceforge.net/

But why don't you join them in the avs script ? e.g. export your credits in a lossless format, join with the script before encoding with x264

2.Uhh, really before upscaling.... can I really finetune that far for those few specific sections :/
Where can I manipulate the TIVTC filter in MeGUI o.o??

And yes, can you give me link to an EDI dll plugin for MeGUI.
My MeGUI doesn't seem to open the filter when I call them...

I think in the edit tab you can edit scripts and save them

TIVTC has many, many settings, you have to look at the instructions

NNEDI3 can be found on tritical's page
http://web.missouri.edu/~kes25c/

e.g. to do upscale, use fwidth=x , fheight=y

NNEDI3_rpow2(2, cshift="spline36resize", fwidth=x, fheight=y)


Looking at your screenshot, it looks like just poor line quality to me . Some other samples didn't have this, so it depends how picky about quality you are and how far you are willing to go to do a better job

[Weef]

1. Would be an idea I haven't considered so far....but got to look up where they put the prolog cuz I have to redo tht too and for that I need a timeline

2. I know it has many settings...but far too many and if I want to have it finished by next month :/

Sure, the poor line quality, that's why I wanted to stay really close to the original dimensions in the first place :/
But in the moving picture they actually look fine but I never now how it looks with other characters.... each one has a different color with a different colored outline...some stay strong...some that have a simillar color next to them turn weaker in upscaling :/

[poisondeathray]

Sure, the poor line quality, that's why I wanted to stay really close to the original dimensions in the first place :/

I hope you realize that when your player or TV scales it, the same artifacts will be revealed . Just by scaling (or not scaling) doesn't affect this (the point is moot) . At least if you choose to address the problems, then you can improve them

But in the moving picture they actually look fine but I never now how it looks with other characters.... each one has a different color with a different colored outline...some stay strong...some that have a simillar color next to them turn weaker in upscaling :/

Not sure what you mean by this, but there are other problems in the samples you posted - e.g. line chroma noise (the black lines aren't pure black) . If I were doing this, I would add some other filters like mild denoisers, chroma denoisers, sharpeners

[Weef]

1. good point

2. Got some tools to make this visible
But I'm not really sure if ANY character in this anime ever had a pure black outline!

[poisondeathray]

2. Got some tools to make this visible
But I'm not really sure if ANY character in this anime ever had a pure black outline!

Maybe not pure black, but I'm referring to "rainbows" that probably shouldn't be there

Just zoom in (e.g upscale) without any other filters

eg. this was from one of samples, just resized with spline36resize(1280,720), and cropped

Look at the "black" lines around the nose, chin, etc... have multicolors . At least I dont' think they should be there. Who knows, maybe he's a "colorful guy" :P

[Weef]

Phew...no sry, barely a little cyan and maybe red next to it...I get the point.
But is the source then?
How is that called?
What can one do about it?

[poisondeathray]

Phew...no sry, barely a little cyan and maybe red next to it...I get the point.
But is the source then?
How is that called?
What can one do about it?

This is in the source. I would call this rainbows or chroma noise.

If you didn't like it you would apply a derainbow filter or chroma noise filter

e.g
fft3filter(plane=3, sigma=3, bt=2)

Of course, you can adjust the strength to remove more or reduce it to remove less . I only reduced it a little here, but you can increase sigma to increase the effect. If you raise it too much, the blue hue will disappear from the lines, and they will become more "black". Maybe you want that, maybe you dont....
You can add other filters, etc.... whatever you want





Generally when you upscale anime, you add some mild denoising, upscale, then some mild sharpening

[Weef]

Ok, if that's the order from your experience that makes sense, then I will use and apply them this way.

[Weef]

Little side question
Are there any results if first changing height then length or vise versa results in different image quality?
Just curious, can't say I would know the math behind it.
Are there tests like that?

[poisondeathray]

Little side question
Are there any results if first changing height then length or vise versa results in different image quality?
Just curious, can't say I would know the math behind it.
Are there tests like that?

What do you mean ?

Something like

Spline36Resize(somewidth, someheight)

vs

Spline36Resize(somewidth)
Spline36Resize(someheight)

?

I don't think there would be any difference. The 2nd one might be slightly slower

[jagabo]

From the AviSynth manual:

AviSynth has completely separate vertical and horizontal resizers. If input is the same as output on one axis, that resizer will be skipped. Which one is called first, is determined by which one has the smallest downscale ratio. This is done to preserve maximum quality, so the second resizer has the best possible picture to work with.

http://avisynth.org/mediawiki/BicubicResize