New RIFE filter - 3x faster AI interpolation possible in SVP!!! (Page 105 of 105)

The v2 models cause stutter for me. I get reduced GPU usage, however there is stutter in the playback without any dropped frames. When I switch back to the v1 models (e.g. 4.25_lite), the stutter goes away.

I've got a 4080 doing 4k from 23.976 fps to 48 Hz.

I am having an issue trying to use the new VapourSynth R78 candidate release with SVP.Previously, when testing versions (like up to R77), I would simply replace the vapoursynth.dll file as usual, and everything worked perfectly. However, after doing the exact same thing with the new R78 DLL, my video player just  closes instantly whenever I try to open a video.If I revert back to the R77 DLL, the player works normal again.(any player)

I know that R78 introduced a breaking change or an API update that causes SVP or the player script to crash.
Does anyone know a way to make R78 work with SVP manually, and what exactly do I need to change? Is there another file or configuration I need to modify.

Thanks in advance for any help!

I'd suggest trying 4.22 Lite and 4.25 Heavy. Both are good quality-wise but while 4.25 Heavy is slightly better, it comes at a very high performance cost. 4.22 Lite has become my default.
With 4.26 models, I've experienced some pretty bad artifacts, so would not recommend. YMMV.

Nevermind, asked Fable and it did some tests:

The setup: RTX 5090 Laptop GPU, SVP's bundled stack (TensorRT 10.8, vstrt.dll) vs TensorRT-RTX 1.6.1 (vs-mlrt v16.2.test1 vstrt_rtx.dll), RIFE 4.26 Heavy, fp16 + static shape + CUDA graph + 2 streams (SVP defaults), synthetic source through vspipe, so the numbers are max interpolation throughput and source fps does not change them. Numbers below are TensorRT vs TensorRT-RTX, median of 5 runs per config; the dashed marks in the chart show what a 23.976 movie needs at each multiplier:

1080p (1920x1088): x2 132.6 vs 135.4, x3 102.8 vs 101.1, x4 91.2 vs 90.3, x5 84.7 vs 84.1. Identical within run to run noise.
4K (3840x2176): x2 30.6 vs 33.0, x3 22.6 vs 24.4, x4 20.3 vs 21.4, x5 19.4 vs 20.4. TensorRT-RTX is 5-8% faster and won every single repetition.
For a 23.976 movie that means 4.26 Heavy is real-time at 1080p up to x3 (x4 barely misses, 91 vs 96 needed) and not real-time at 4K even at x2.

Engine build for 4.26 Heavy: 99 to 112 seconds on TensorRT 10.8 vs 0.5 seconds on TensorRT-RTX (about 7 seconds total setup including first engine load). Rebuilt engines from scratch twice per backend to confirm, it holds.

Also confirmed the weird-size crash is really fixed in 1.6.1: 1920x832, 1280x768 and 896x512 all build and render fine.

So: same speed at 1080p, faster at 4K, and the 100 second engine build basically disappears. Looks like a clear win. SVP's script generator already has the rife_rtx hook that loads vstrt_rtx.dll, it just needs the updated rife package (vstrt_rtx.dll + tensorrt_rtx DLLs + newer vsmlrt.py).

How can I properly use rife_rtx?
Here is what I have done so far:
Placed tensorrt_onnxparser_rtx_1_6.dll, tensorrt_rtx_1_6.dll, and tensorrt_rtx.exe into SVP 4\rife\vsmlrt-cuda.
Placed vstrt_rtx.dll and the updated vsmlrt.py into SVP 4\rife.
However, it doesn't seem to work: SVP still uses the regular TensorRT backend (resulting in slow engine compilation) instead of TensorRT-RTX.
Could someone explain how to correctly trigger or activate rife_rtx in SVP 4? Thanks!

If you want to apply TensorRT-RTX, do this:

https://drive.google.com/drive/folders/ … XX-S2JMOXD

1. Close SVP.
2. Make a backup of your "SVP 4" folder.
3. Copy the "SVP 4" folder from the archive on top of yours, overwrite all.
4. Start SVP and use RIFE as usual.

Nothing else to configure. The first time a model or a resolution is used
it prepares the engine by itself, after that it is cached.

Numbers here, RTX 5090 Laptop, model 4.6, x2, 5 runs each:

                      TensorRT      TensorRT-RTX
1080p                261.8 fps       267.4 fps
4K                    60.0 fps        62.8 fps
engine build 1080p       103 s            12 s
engine build 4K          122 s            14 s
seek in mpv              0.9 s          0.45 s
runtime files          2565 MB          222 MB

The last row is what the two backends need on disk to build and run engines.
The archive does not remove anything from your install, it adds the RTX
runtime next to what you already have, but it shows what SVP could ship
instead of the old TensorRT 10 builder libraries.

Thanks for response! But I've got error after replacing files.

> With 1.6.1 the odd-size crash is fixed and the speed is 261.8 vs 267.4 fps at 1080p and 60.0 vs 62.8 at 4K (model 4.6, x2).

is it possible that 5090 is not a bottleneck in your case?
what is the GPU load?