Do you have charted out performance of different models

[dnotario]

Same way you have the quality?

[hzwer]

No, I only have this.

[NevermindNilas]

https://github.com/styler00dollar/VSGAN-tensorrt-docker#:~:text=Rife4%2Bvs%20(ensemble%20False)

There are a couple of performance benchmarks here that go up to 4.12-lite but these use vapoursynth so they are a tiny bit faster than the 'pure' python implementations.

But for whatever it's worth, with a similar implementation as the one in this repo I get about 77/78 FPS with Rife 4.15/4.14/4.14-lite for 1080p inputs and 2x interpolation and a RTX 3090.

[banjaminicc]

Makes> https://github.com/styler00dollar/VSGAN-tensorrt-docker#:~:text=Rife4%2Bvs%20(ensemble%20False)

There are a couple of performance benchmarks here that go up to 4.12-lite but these use vapoursynth so they are a tiny bit faster than the 'pure' python implementations.

But for whatever it's worth, with a similar implementation as the one in this repo I get about 77/78 FPS with Rife 4.15/4.14/4.14-lite for 1080p inputs and 2x interpolation and a RTX 3090.

It's not vapoursynth that makes the implementation faster. It's tensorrt

[NevermindNilas]

Sure.