Selur's Little Message Board

Full Version: increase panning smoothness
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2
I'm crying! 
Thank you for your help! 
It all worked. 
You're a saint!!!
Is it possible to somehow link this ccvfi script with FillDuplicateFrames?
Currently FillDuplicateFrames can only use SVP or RIFE, which give bad results compared to this ccvfi script.

p.s. FillDuplicateFrames has an incorrect description, it is from ReplaceBlackFrames.
Quote: Is it possible to somehow link this ccvfi script with FillDuplicateFrames?
Sure, by expanding the script.

Quote:FillDuplicateFrames has an incorrect description, it is from ReplaceBlackFrames.
What's wrong with:
Code:
'''
call using:

from FillDuplicateFrames import FillDuplicateFrames
fdf = FillDuplicateFrames(clip, debug=True, thresh=0.001, method='SVP')
//fdf = FillDuplicateFrames(clip, debug=True, thresh=0.001, method='MV')
//fdf = FillDuplicateFrames(clip, debug=True, thresh=0.001, method='RIFE')
clip = fdf.out

Replaces duplicate frames with interpolations.
v0.0.3
0.0.4 removed and added back RGBH support or RIFE
0.0.3 allow to set device_index for RIFE and support RGBH input for RIFE
0.0.4 removed RGBH since RIFE ncnn does not support it
0.0.5 add general sceneThr
0.0.6 add rifeModel parameter
0.0.7 add mode: FillDuplicate|FillDrops|Replace, add: rifeTTA, rifeUHD
'''
?
I meant this description:
[Image: _8ea7115ea8d6539c3709f80f5b654f4c.jpeg]

I tried to do it like this, but it doesn't work Sad
[Image: 8c373813f5eb4b226486ce458e08bc72.jpeg]
Ah, will fix the tool-tip.

I thought that:
https://pastebin.com/ZH6aPUE2
with:
Code:
# requires colorformat RGBH

from FillDuplicateFrames import FillDuplicateFrames
fdf = FillDuplicateFrames(clip, debug=True, thresh=0.001, method='CCVFI')
clip = fdf.out
would work, but it does not.
It fails with:
Code:
File "F:\Hybrid\64bit\Vapoursynth\Lib\site-packages\ccvfi\vs\vfi.py", line 44, in inference_vfi
..
ValueError: Clip do not have enough frames for vfi method require 3 frames once infer
Looking at vfi.py, I'm not sure how to 'fix' that. Sad
(FillDuplicateFrames interpolates between 2 frames,...)

Cu Selur
Quote:(FillDuplicateFrames interpolates between 2 frames,...)

But ccvfi also interpolates between two frames, right?
Quote:in_frame_count: int = 2

But the output should be three frames, two original and one new, right?
No clue, how ccvfi works internally.
Might be that it looks at more than 2 frames.
Number of output frames depends on the target fps, which here is the same as the target frame rate.
=> I don't know enough about ccvfi to adjust FillDuplicateFrames for it (assuming it's possible).

Cu Selur
It seems that everything is really complicated.
I conducted a small experiment, ccvfi really refused to work on three frames, it starts working starting from a certain number of frames.
And the algorithm of its work is not so simple.
Sample:
https://gofile.me/7kQw9/YRjllVqtn

Is it possible then to add the RIFE model selection to the interface in FillDuplicateFrames?
It is not very convenient to select while editing the script.
I probably can add a model selection, but be warned that this might cause problems, see: https://github.com/styler00dollar/Vapour.../issues/16 => so I'll limit the model selection (and RIFE version).

Cu Selur

Ps.: uploaded a new dev with RIFE model selection for FillDuplicateFrames.
Pages: 1 2