Home Artists Posts Import Register

Content

Lots of change, hoping it don't have a lot for new bugs as well.

  • New options to interpolate X16 and X32, why not?
  • Concat option create a new video instead of replacing the default output.
  • You can select a PNG image as input, the app will scan the PNG folder for all PNG files and order them by name. I'm sure this workflow can be improved in the future.
  • You can export PNG sequence, the mp4 is still created with this option enable.
  • [1.] You can select the CRF to change the quality of the output MP4
  • [2.] There is a few models to select, they change the output result. Personally I think that 2.3.1 work pretty good for most animations, but you have to test each model to see what work best for your input. Previous version of Rife-App use the model 2.4
  • [3.] To help test the output, you can select to just render a few frames of your input.

[4.] This is a experimental option, here a example, an animation using this 3 frames with 32X:

Doing a regular interpolation X32, this is the result:

And here the result with "Smooth Interpolation"

The idea is to give a little more energy to animations with few frames, like gifs.

Files

RIFE-App 1.72.7z

Comments

Anonymous

Will you implement a mode 3 in the future? I think that would best make use of x16 and x32 without being wasteful

Snake Plissken

I've heard there is such a thing as variable frame rate video? (alternative is to use frame doubling in some areas) Could possibly reduce workload by a lot of the user maybe includes a txt file or someother formate specifying the frames / timestamps so that only specific areas of the video get interpolation. (less work than manually cutting video up into regions)

DAINAPP

Variable framerate are only useful exporting as gif and perhaps some other types. I not sure I understand the txt file suggestion.

Anonymous

Hi there. I got a strange problem here with my device. I got two same graphic gards, it can only run on the first one. When I try to run on the second card "RuntimeError: Expected tensor for argument #1 'input' to have the same device as tensor for argument #2 'weight'; but device 0 does not equal 1 (while checking arguments for cudnn_convolution)" and stop there.

DAINAPP

Thanks for the report, I think I know why this bug is happening, will fix it.

Anonymous

Hey sorry for the dumb question, is this a different app than the DAIN APP?

Anonymous

Yes, this is a newer version of it running with the RIFE algorithm wich is way faster and has less artifacts etc. just way better in pretty much any point. And RIFE is getting updates whereas DAIN doesnt anymore. Hope i could help :)

Anonymous

Thanks so much, I've been using DAIN this whole time lol.

Anonymous

Really cool software. Just curious, is there a way to export the subframes with decimal places? So instead of 001, 002, 003, we get 001, 001.25, 001.50, 001.75, 002, 002.25 and so on?

Anonymous

That would be great if easy to do. This would allow those frames to be read as sub-frames in other software. Maybe quite niche!

Anonymous

So what exactly are the differences between the models? I can kinda see some differences but I'm not entirely sure what I'm looking at.