I need some help with interpolation videos, lots of detail in post

[lansing]

People need to actually watch op's video before giving opinions. The video he's talking about is a cgi movie, not animation, there are no repeated frames, so everything you said about animation is invalid to him.

[lansing]

Originally Posted by FranceBB

Blending doesn't look as smooth as motion interpolation, but it doesn't make the video stutter as a repeated frame does.

No, blending is visually the same thing as repeated frame. There's no new interpolated object added in between, so on playback you're still going to see the same 2 frames, same old stuttering.

[bradwiggo]

Originally Posted by bxyhxyh

As johnmeyer said
This nature of animation leads to ugly results instead of getting smoother video.

For example,
Think that there are 2 objects move in the video.
They don't move together sometimes. They might even animated separately.
One moves at 12 fps and other one is 8 fps etc...
Not full 24 fps.

Any interpolation isn't smart enough to see this.
They will just add duplicate of that object since it can't see objects as 'moving'.
Result wouldn't be not much smoother than the original if not any smoother.

That's why he is saying it's impossible.
But you can try.

I know it isn't impossible though, as I have found a video of it. My current goal is simply to reproduce the linked youtube video.

[bradwiggo]

Originally Posted by poisondeathray

One difference is the YT version you linked to uses scene blending . That could explain the "stutter" that you are seeing, because there would be some duplicate frames at the end of each scene. So in any of those generic interpolation scripts, you would use blend=true for the flow part . Usually it's set to false by default for most interpolation scripts for any generic scripts

In my experience, GPU quality is noticably worse for interpolation for SVP ; larger artifacts, and sometimes even duplicate frames. I posted examples and comparisons probably here and other forums as well. This might not be true for your system or card/hardware setup, so try different combinations . But it's usually much faster

Disable all the other options , things like like artifact masking if you want it to look like the YT version

If you do all that and still think the YT version is "smoother" ; the other possibility is some playback issue; there might be differences in HW acceleration in browser or local media playback.

Which script would you recommend using in order to reproduce the video, would it be one of the ones posted by people here?

I don't think it will be a playback issue, as I have downloaded the youtube video using the link in the video description that actually links to the file, and I played it in my normal media player and it still looks better.

[johnmeyer]

Originally Posted by lansing

People need to actually watch op's video before giving opinions. The video he's talking about is a cgi movie, not animation, there are no repeated frames, so everything you said about animation is invalid to him.

Yes, if you read my last post, I acknowledged that. No need for this post.

[johnmeyer]

Originally Posted by lansing

No, blending is visually the same thing as repeated frame. There's no new interpolated object added in between, so on playback you're still going to see the same 2 frames, same old stuttering.

There is no stuttering. The visual artifacts of the original movie are simply those which happen with 24 fps progressive material. It has been known for 100+ years -- going back to 12-16 fps hand-cranked movies -- that you get "judder," a visual disturbance that is entirely created within your head because these lower frame rates -- including the universal sound film 24 fps speed -- is lower than the threshold for human persistence of vision.

Thus, the OP's original desire to increase the frame rate in order to eliminate these visual disturbances is quite well founded, but the reality that he won't seem to acknowledge is that the technology does not exist to do this on all scenes. For a month he has posted that he thinks this is possible because he has seen examples where 24 fps has been increased in frame rate without introducing motion estimation artifacts. The problem is, these examples show scenes where ME works just fine, but it will always fail on scenes with attributes that I have described multiple times in previous posts.

Of course if you can come up with a solution for his video that works, my hat is off to you!

[johnmeyer]

Originally Posted by bradwiggo

I think I should have worded my original post better, as I was not necessarily looking for a perfect script for the entire film, I understand that is most likely not possible, I was instead looking for the script that was used to make that youtube video.

If you re-read my post, I was responding to Lansing, not you (i.e., not the name in the quote in my post).

[poisondeathray]

I just copy and pasted the "Manolito mod" posted here where you said it had stutter and indeed there were duplicate frames. I explained why (or at least one of the reasons why) - the scene changes have duplicated frames when you set blend=false. It's the same for SVPFlow or MFlowFPS or any of the avisynth interpolation functions.

You said the YT video used SVPFlow, then use SVPFlow . There must be some combination of settings that reproduces it, but 100% certain it uses blend=true (for the scene change) . If you go frame by frame in the YT you will see this. Although in the comments the guy wasn't sure what was used...

If you can't explain in words why one is "smoother", then compare it frame by frame

And you don't need to encode a video to preview it, you can preview it in avspmod or vdub2 . Go frame by frame or even stackhorizontal() with the youtube video (resize either yours to 568 ,or YT's to 570 height)

Originally Posted by lansing

No, blending is visually the same thing as repeated frame. There's no new interpolated object added in between, so on playback you're still going to see the same 2 frames, same old stuttering.

Not really. Visually they are different. Blending is slightly smoother but gives a "strobey" or "ghosting" look. Repeated frames is the cleanest, but the least smooth - it's what most people would say exhibits the most "stutter"

For example try ChangeFPS vs. ConvertFPS . You're implying they give visually same result ?? They definitely don't.

[bradwiggo]

Originally Posted by johnmeyer

There is no stuttering. The visual artifacts of the original movie are simply those which happen with 24 fps progressive material. It has been known for 100+ years -- going back to 12-16 fps hand-cranked movies -- that you get "judder," a visual disturbance that is entirely created within your head because these lower frame rates -- including the universal sound film 24 fps speed -- is lower than the threshold for human persistence of vision.

Thus, the OP's original desire to increase the frame rate in order to eliminate these visual disturbances is quite well founded, but the reality that he won't seem to acknowledge is that the technology does not exist to do this on all scenes. For a month he has posted that he thinks this is possible because he has seen examples where 24 fps has been increased in frame rate without introducing motion estimation artifacts. The problem is, these examples show scenes where ME works just fine, but it will always fail on scenes with attributes that I have described multiple times in previous posts.

Of course if you can come up with a solution for his video that works, my hat is off to you!

My original post was not clear enough on what I was trying to achieve. I do understand that a script that makes the whole movie look as good as that video does not exist (or at least is very very unlikely to exist), however, that is not currently my aim. What I am currently trying to achieve is to find the script that made that youtube video, regardless of how well that script would work with the rest of the film.

[poisondeathray]

This comparison is aligned (using trim()) and stacked; each is resized 1212x540 to keep approx. AR (the reason is so it can be stacked and viewed on a 1080 height screen).

They are not labelled on purpose - Which one is the YT video, which one is the jm_fps manolito mod (with blend=true, masking disabled) using your test1.mkv ? It's easy to tell from a compression standpoint (the YT version will have more compression artifacts)
http://www.mediafire.com/file/vmi3pm...mpare.mp4/file

Some frames slightly better, some slightly worse, but in terms of overall "smoothness" , I'd argue it's fairly close. I'm sure you can tweak the settings a bit to make it even better in some scenes, but the main differences between typical scripts is the blend=true . (Most of the time people don't want blending for general use scenarios)

I understand you're mainly interested in "smoothness" only here, less so about artifacts . But artifacts can contribute to the perception of reduced smoothness; so don't automatically discount artifacts either. But clearly both have ugly artifacts, some better some worse

OR - if you still think one is more smooth, then identify which one and why, or what about it is more smooth ?

[Sparktank]

Originally Posted by bradwiggo

What I am currently trying to achieve is to find the script that made that youtube video, regardless of how well that script would work with the rest of the film.

An impossible feat.
From the comments in the youtube video, the uploader doesn't know anything either. He just found it and uploaded it.

But, given that it was in 2014, the options were limited back then.
SVP/Spriton/mvtools.

Your best bet is to stick with the scripts posted.

There was probably some light artifact masking in it if they used SVP Pro.
Free vs Pro, you get to customize a lot in Pro (SVP 4).
back then, I think it was only SVP3 and still customizable before it went SVP 4 Pro.

You're not going to be able to 'guess' the settings.
If you can find the original person who created the video, they might still have the script.
Or they just might say use SVP.

But I tried the jm_fps script and the mod the other day on a ProRes trailer (1080p) and it came out incredibly smooth without much artefacts (except for a few complex scenes).


Your best bet is to play with SVP and its settings (if you paid for it).
The SVPFlow library uses some of MVtools settings, which you can read about here:
https://www.svp-team.com/wiki/Manual:SVPflow

With SVP 4 Pro, using the max settings for everything (encoding/playback) doesn't always guarantee "the best" settings.

There were days where I spent a couple hours going through various scenes before settling on an average to finally watch the whole movie.

[lansing]

Originally Posted by poisondeathray

Not really. Visually they are different. Blending is slightly smoother but gives a "strobey" or "ghosting" look. Repeated frames is the cleanest, but the least smooth - it's what most people would say exhibits the most "stutter"

For example try ChangeFPS vs. ConvertFPS . You're implying they give visually same result ?? They definitely don't.

Yes you are right about the ghosting with blending frames, but that does not make it smoother. When an object is moving from position A to position B, with the ghosting effect, visually you kind of see the object arrived to B before it actually happened, but the motion of the object does not get smoother, it still would be stuttering.

[poisondeathray]

Originally Posted by lansing

Yes you are right about the ghosting with blending frames, but that does not make it smoother. When an object is moving from position A to position B, with the ghosting effect, visually you kind of see the object arrived to B before it actually happened, but the motion of the object does not get smoother, it still would be stuttering.

It's usually distinguished from "stuttering" which implies a pure repeat frame (a cadence of repeat frames , usually irregular , like AABBBCDDDDDD)

People usually call "blending" slightly smoother than pure repeats because edges are less aliased. It's along the same lines as motion blur the makes the appearance of motion smoother. The object hasn't really moved their either

Personally I think blending is terrible, but ask 10 people and 9/10 people will say it's slightly smoother than pure repeats . Personally I find the "strobey" look nauseating

[bradwiggo]

Originally Posted by Sparktank

An impossible feat.
From the comments in the youtube video, the uploader doesn't know anything either. He just found it and uploaded it.

But, given that it was in 2014, the options were limited back then.
SVP/Spriton/mvtools.

Your best bet is to stick with the scripts posted.

There was probably some light artifact masking in it if they used SVP Pro.
Free vs Pro, you get to customize a lot in Pro (SVP 4).
back then, I think it was only SVP3 and still customizable before it went SVP 4 Pro.

You're not going to be able to 'guess' the settings.
If you can find the original person who created the video, they might still have the script.
Or they just might say use SVP.

But I tried the jm_fps script and the mod the other day on a ProRes trailer (1080p) and it came out incredibly smooth without much artefacts (except for a few complex scenes).


Your best bet is to play with SVP and its settings (if you paid for it).
The SVPFlow library uses some of MVtools settings, which you can read about here:
https://www.svp-team.com/wiki/Manual:SVPflow

With SVP 4 Pro, using the max settings for everything (encoding/playback) doesn't always guarantee "the best" settings.

There were days where I spent a couple hours going through various scenes before settling on an average to finally watch the whole movie.

Would the interpolation using SVP 3 look better than using SVP 4? Or if not necessarily better would there be a significant difference.

[lansing]

Originally Posted by johnmeyer

Thus, the OP's original desire to increase the frame rate in order to eliminate these visual disturbances is quite well founded, but the reality that he won't seem to acknowledge is that the technology does not exist to do this on all scenes. For a month he has posted that he thinks this is possible because he has seen examples where 24 fps has been increased in frame rate without introducing motion estimation artifacts. The problem is, these examples show scenes where ME works just fine, but it will always fail on scenes with attributes that I have described multiple times in previous posts.

Of course if you can come up with a solution for his video that works, my hat is off to you!

I thought he was only asking about how to get the result of the youtube video, which can easily be achieved by using the sample script provided by the svpflow package.

The only thing he needs to do now is to figure out how to get his amd intergrated graphic card to work with the script.

[Sparktank]

Originally Posted by bradwiggo

Would the interpolation using SVP 3 look better than using SVP 4? Or if not necessarily better would there be a significant difference.

They made a lot of improvements to 4, so I'd say 3 would be worse.
I did some testing after 4 came out and never went back to 3.

It's $24.99 (USD).
https://www.svp-team.com/wiki/Purchase

you can do just as good with the scripts posted in the forum.
you're not going to find better out there (or here).

Unless something significant happens like a whole new level to the mvtools code.
Which seems to be mostly just updates for avs+ and new-age colorspaces (beyond yv12/yv16/yv24).

Or wait until NVidia becomes more popular with it's cuDNN library and find a way into mainstream editors.
https://forum.doom9.org/showthread.p...68#post1845068

[poisondeathray]

Originally Posted by poisondeathray

This comparison is aligned (using trim()) and stacked; each is resized 1212x540 to keep approx. AR (the reason is so it can be stacked and viewed on a 1080 height screen).

They are not labelled on purpose - Which one is the YT video, which one is the jm_fps manolito mod (with blend=true, masking disabled) using your test1.mkv ? It's easy to tell from a compression standpoint (the YT version will have more compression artifacts)
http://www.mediafire.com/file/vmi3pm...mpare.mp4/file

And here is the same thing, just pillarboxed to get 1080p60 treatment on YT
https://www.youtube.com/watch?v=DUmHTLkEgec

Again, tell me which is which. Or which do you think is "smoother" ?

[bradwiggo]

Originally Posted by poisondeathray

And here is the same thing, just pillarboxed to get 1080p60 treatment on YT
https://www.youtube.com/watch?v=DUmHTLkEgec

Again, tell me which is which. Or which do you think is "smoother" ?

So is one of those the file I uploaded and one is the youtube video? Could you upload those 2 videos but separately (but with random names so I don't know which is which), as I struggle to see the fine details when it is only half the size of the screen (my laptop only has a 720p screen).

[poisondeathray]

Originally Posted by bradwiggo

So is one of those the file I uploaded and one is the youtube video? Could you upload those 2 videos but separately (but with random names so I don't know which is which), as I struggle to see the fine details when it is only half the size of the screen (my laptop only has a 720p screen).

Yes; but it will be easy to tell from the filesize difference

The "smoothness" is pretty much the same. There is nothing "special" about the youtube video. The only significant difference between any default script setting is the blended scene changes

When you stack them and use the same player, you eliminate all the other potential issues maybe you had a source decoding issue introducing duplicates ; or playback issues like different decoder, different player, different renderer. (Even the same player can use different decoding pathway)




You can download the YT version, and make the other one yourself. As mentioned earlier, the script is the jm_fps manolito mod (I just copied and pasted from your earlier post), with scene blending enabled, masking disabled

You can go frame by frame , or if you want compare them in different tabs in avspmod (if you resize to same dimensions, trim() to align the frames - they will be superimposed and you just hit the number keys to swap tabs back and forth - very easy to see frame differences this way)

Or if you're just looking at individual videos, separately, then you don't need to resize or align them. But it's easy to see they they are very similar in terms of "smoothness".

Or, if you still think they are different in terms of "smoothness" then we need to investigate farther - eg. maybe there was an encoding issue your end?, maybe the encoding settings you're using are causing problems? YT videos are encoded so they can be easily decoded across platforms. Maybe player dropping frames, etc...

FFVideoSource("test1.mkv")
jm_fps(59.94)

As mentioned earlier, only the very last line is changed (highlighted in red) ; to blend=true, and the masking disabled (rest of line commented out). You could do the same thing in SVPFlow or any of the dozen interpolation variants. They produce similar results when using similar settings

# Motion Protected FPS converter script by johnmeyer from Doom9
# Slightly modified interface by manolito
# Requires MVTools V2 and RemoveGrain
# Also needs fftw3.dll in the System32 or SysWOW64 folder for Dct values other than 0


function jm_fps(clip source, float "fps", int "BlkSize", int "Dct")
{
fps = default(fps, 25.000)
fps_num = int(fps * 1000)
fps_den = 1000
BlkSize = default(BlkSize, 16)
Dct = default(Dct, 0)

prefiltered = RemoveGrain(source, 22)
super = MSuper(source, hpad = 16, vpad = 16, levels = 1, sharp = 1, rfilter = 4) # one level is enough for MRecalculate
superfilt = MSuper(prefiltered, hpad = 16, vpad = 16, sharp = 1, rfilter = 4) # all levels for MAnalyse
backward = MAnalyse(superfilt, isb = true, blksize = BlkSize, overlap = 4, search = 3, dct = Dct)
forward = MAnalyse(superfilt, isb = false, blksize = BlkSize, overlap = 4, search = 3, dct = Dct)
forward_re = MRecalculate(super, forward, blksize = 8, overlap = 2, thSAD = 100)
backward_re = MRecalculate(super, backward, blksize = 8, overlap = 2, thSAD = 100)
out = MFlowFps(source, super, backward_re, forward_re, num = fps_num, den = fps_den, blend = true)#, ml = 200, mask = 2)

return out
}

[bradwiggo]

Originally Posted by poisondeathray

Yes; but it will be easy to tell from the filesize difference

The "smoothness" is pretty much the same. There is nothing "special" about the youtube video. The only significant difference between any default script setting is the blended scene changes

When you stack them and use the same player, you eliminate all the other potential issues maybe you had a source decoding issue introducing duplicates ; or playback issues like different decoder, different player, different renderer. (Even the same player can use different decoding pathway)




You can download the YT version, and make the other one yourself. As mentioned earlier, the script is the jm_fps manolito mod (I just copied and pasted from your earlier post), with scene blending enabled, masking disabled

You can go frame by frame , or if you want compare them in different tabs in avspmod (if you resize to same dimensions, trim() to align the frames - they will be superimposed and you just hit the number keys to swap tabs back and forth - very easy to see frame differences this way)

Or if you're just looking at individual videos, separately, then you don't need to resize or align them. But it's easy to see they they are very similar in terms of "smoothness".

Or, if you still think they are different in terms of "smoothness" then we need to investigate farther - eg. maybe there was an encoding issue your end?, maybe the encoding settings you're using are causing problems? YT videos are encoded so they can be easily decoded across platforms. Maybe player dropping frames, etc...

FFVideoSource("test1.mkv")
jm_fps(59.94)

As mentioned earlier, only the very last line is changed (highlighted in red) ; to blend=true, and the masking disabled (rest of line commented out). You could do the same thing in SVPFlow or any of the dozen interpolation variants. They produce similar results when using similar settings

# Motion Protected FPS converter script by johnmeyer from Doom9
# Slightly modified interface by manolito
# Requires MVTools V2 and RemoveGrain
# Also needs fftw3.dll in the System32 or SysWOW64 folder for Dct values other than 0


function jm_fps(clip source, float "fps", int "BlkSize", int "Dct")
{
fps = default(fps, 25.000)
fps_num = int(fps * 1000)
fps_den = 1000
BlkSize = default(BlkSize, 16)
Dct = default(Dct, 0)

prefiltered = RemoveGrain(source, 22)
super = MSuper(source, hpad = 16, vpad = 16, levels = 1, sharp = 1, rfilter = 4) # one level is enough for MRecalculate
superfilt = MSuper(prefiltered, hpad = 16, vpad = 16, sharp = 1, rfilter = 4) # all levels for MAnalyse
backward = MAnalyse(superfilt, isb = true, blksize = BlkSize, overlap = 4, search = 3, dct = Dct)
forward = MAnalyse(superfilt, isb = false, blksize = BlkSize, overlap = 4, search = 3, dct = Dct)
forward_re = MRecalculate(super, forward, blksize = 8, overlap = 2, thSAD = 100)
backward_re = MRecalculate(super, backward, blksize = 8, overlap = 2, thSAD = 100)
out = MFlowFps(source, super, backward_re, forward_re, num = fps_num, den = fps_den, blend = true)#, ml = 200, mask = 2)

return out
}

Did you cut the youtube video to the same size as the one I made? I will try cutting it and then go through frame by frame with the youtube one and my one on a different monitor.