1 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

cemaydnlar wrote:
Chainik wrote:
cemaydnlar wrote:

Can i get a link for rife 4.12 and a quick guide on how to use it on svp pls?

https://github.com/AmusementClub/vs-mlr … nal-models
1. put .onnx into rife models\rife folder
2. restart SVP


Love u man. Thx now another question for those who tried 4.13 lite and 4.12v2 which one is better for anime ?

4.12 is best for animation, the lite model is slightly worse than 4.12 but performs better.

Go to forum: Using SVP

2 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

cemaydnlar wrote:

I am using rife 4.9 x3 with vsync and it creates a lot of artifacts on spy x family anime. Did you guys try this anime. I thought 4.9 should be better in the frame of artifacts but if i'm not mistaken it is the same as 4.6 with poor gpu performance.

For anime with a lot of hardcoded subtitles, try a model >= 4.10, even 4.12_lite is much better than 4.9.

Go to forum: Using SVP

3 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

Blackfyre wrote:
zerosoul9901 wrote:

I tried RIFE 4.10 and thought it greatly improved hard coded subs, with previous versions of RIFE the letters mostly get warped to a point when encountering fast motion scenes, now with RIFE 4.10 it's a lot more stable.

Tested and can confirm my RTX 3090 is not strong enough to be able to push 4K @ 2x

Back to 4.9 for me, so far the best results, but still very demanding 99% GPU usage, so RTX 3090 overclocked is just barely enough for 4K @ 2x

For Hardcoded Subtitles that are permanently embedded into a video, 4.9 is even worse than 4.8, more Distorted/Warped Subtitles in dynamic scenes, 4.10 is the most stable version i've tested even though the gpu usage is the heaviest

For watching Anime in 1080p with a lot of Hardcoded Subtitles, 4.10 is the best.

Go to forum: Using SVP

4 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

I tried RIFE 4.10 and thought it greatly improved hard coded subs, with previous versions of RIFE the letters mostly get warped to a point when encountering fast motion scenes, now with RIFE 4.10 it's a lot more stable.

Go to forum: Using SVP

5 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

dlr5668 wrote:
Chainik wrote:

> while v2 models significantly reduces the playback seek time, improve performance

dunno, I can't see any improvements
not "significantly", not even slightly

----
unpack "rife_v2" into SVP 4\rife\models (i.e. there must be SVP 4\rife\models\rife_v2\rife_v4.6.onnx)
replace SVP 4\rife\vsmlrt.py

switch back to "v1" models - rename "rife_v2" folder to any other name, like "_rife_v2"

v2 takes 7.5/8 and crashes at seek, v1 stays at 3.5/8 as it should

Don't use dynamic shapes, it doesn't seem to work well with v2 models and requires more VRAM usage compared to v1, use static shapes instead, and use flag force_fp16 instead of fp16 during inference, reducing memory usage and engine build time during engine build, you can also add _implementation=2 flag to easily switch v2 models.

Go to forum: Using SVP

6 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

UHD wrote:

Thanks Mardon85 for the further tests and I'm very glad you found a way to interpolate x3 in real time. I think the setting you showed might also help someone.

One more thing can increase efficiency:

scripts/vsmlrt.py: added support for rife v2 implementation
(experimental) rife v2 models can be downloaded on https://github.com/AmusementClub/vs-mlr … nal-models ("rife_v2_v{version}.7z"). It leverages onnx's shape tensor to reduce memory transaction from cpu to gpu by 36.4%. It also handles padding internally so explicit padding is not required.

This update came out a few days ago: https://github.com/AmusementClub/vs-mlr … c649dfb212

I guess all you need to do is replace the vsmlrt.py file and download the rife v2 models. And maybe remove explicit padding, I am not a programmer.

Share your impressions on how this modification, and what Mardon85 suggests, affects performance.

From my tests, using ensemble models improves frame interpolation quality, while v2 models significantly reduces the playback seek time, improve performance and should also reduce RAM bandwidth requirements.

Go to forum: Using SVP

7 Reply by zerosoul9901

(1,489 replies, posted in Using SVP)

Unable to download Rife TensorRT when reinstalling Svp4, it seems that there is something wrong with the server.

https://i.imgur.com/ZuDP4ig.jpeg

Go to forum: Using SVP