Rife-App 3.3 - AI video interpolation (x86, x64) (ex DAIN-App) [2022, ENG]

Смотреть в максимальном качестве
https://www.youtube.com/watch?v=0OXzVGLRhK0
https://www.youtube.com/watch?v=q2i6FXVjNT0
С применением Topaz Video Enhance AI:
https://www.youtube.com/watch?v=IjfQnjcrpHo
https://www.youtube.com/watch?v=37D098y1qzM
Создание эффекта slow-motion:
https://youtu.be/60DX2T3zyVo
Старый метод через DAIN-App:
https://www.youtube.com/watch?v=yjMYefRXikI
https://www.youtube.com/watch?v=yLVD58lF0jg

Introduction
There is no configuration on Rife-App that can handle all inputs, even for animation you may
be able to get better results with the default options for one input and totally different options.
This document is a few tips you can try to get better results for your animation.
You should try this setup in order until you manage to get a result you like.
Test 1:
● Turn on “Animation Mode”, animations usually use the same frame for a few
seconds, this helps smooth that out.
● If you have a low quality input, set Similar frame percentage to 0.99 or even 0.98
(instead of 0.999)
● If your animation is a perfect loop, turn on “Perfect loop”
● Set X2 / X4 / X8 depending on what you want.
● If you are 8X or more, you usually will want to turn on “Smooth interpolation” to focus
more on the original frames.
Test 2:
● Turn on “Animation Mode”
● Set Alternative Interpolation 100%
● Alternative Interpolation usually work better with odd multipliers, so test it with 3X /
5X / 7X etc
Test 3:
● If you like the result of Test 2 but there is still to much artifacts, test it with Alternative
Interpolation 75% / 50% / 25%
● If you like the result, but the motion seen floaty, turn on “Smooth Interpolation”
I am doing PixelArt:
● If you are using a really low resolution animation or pixel-art, you can try to put “Inner
scale” to 2.0 or 4.0, it sometimes improves the result.
I don’t want to alter the animation that much:
● You can try to interpolate at X1, this will only create new frames in places where the
animation is completely still, giving a little more flow to the output.

Автор софта выкладывает прогу в виде архива. А они у нас тут запрещены.
Сделал инсталлятор через Inno Setup 6.2 с мощной архивацией. Иконку сделал из постера.

https://grisk.itch.io/rife-app/devlog
Rife-App 2.80
It have new models for animation and RL.
It have an alpha channel option for PNG seq. and gif output
A few minor fixes, like a filename with a dot in its name.
A lot of the memory improvements are kinda of experimental, so let me know if you find any error and I will fix asap.
Rife-App 3.0
Version 3.0 with both Anime and Live action (new name, it's real life model) models.
The idea was to release it after the Anime model was ready (as in the pool result), but I had to re-train it with a few modifications, so it gave me time to finish the Live action model.
This version have pretty much the same options for the last update, plus the new models and bug fixes.
Instead of 4 options per model, there is 3 now. You can calculate the flow using 100% of the resolution, 50% and a fixed aspect ratio. The fixed aspect usually is good for most stuff, but you need to try what is best for each footage.
The fixed aspect bigger advantage is that is fast and it will probably never run out of Vram.
Rife-App 3.10
Pretty much the stuff from the last posts. Some changes in the models, lighter and a little faster.
A new option "Clean Jpeg Artifacts", but is still in alpha, there is a lot to improve on it yet.
Rife-App 3.20
So here is my first attemp with the AMD card support. If anyone have a AMD card please send me some feedback after testing it.
You just need to select "AMD" on the device list.
The models are also slighter better than the last release.
Rife-App 3.30
Pretty much it have new models for animation and Life-Action.

For those newer on this Patreon, this is my app to interpolate frames. Input a anime or video and make more frames for it.
This update took quite a long time to be build. I swear I try to make the next one a loooot faster than this one. There was a lot of learning to make this build, mostly on the Pytorch area.
I can only hope that users of the older versions can find the model to be improved.
Tomorrow I will release a github with some codes that I made to train this models, and then start researching some new methods for interpolation. I feel like I'm pretty much close to the limit of this current method.