You are not logged in. Please login or register.
SmoothVideo Project → Using SVP → New RIFE filter - 3x faster AI interpolation possible in SVP!!!
At least NVOF is still here... I was in doubt because NVOF SDK was not updated since 2023, and that thing was just a techno-demo, not usable in a real life and again, was not updated for two years.
But the NVIDIA guy on their dev forums just replied: "Optical Flow is quite essential to Frame Generation so yes, OF is still supported."
At least NVOF is still here... I was in doubt because NVOF SDK was not updated since 2023, and that thing was just a techno-demo, not usable in a real life and again, was not updated for two years.
But the NVIDIA guy on their dev forums just replied: "Optical Flow is quite essential to Frame Generation so yes, OF is still supported."
Does SVP RIFE TensorRT use Optical Flow or is it only for the other modes?
NVOF is not an "AI" thing at all.
"SVP RIFE TensorRT" can use NVOF for a scene change detection only.
NVOF is not an "AI" thing at all.
"SVP RIFE TensorRT" can use NVOF for a scene change detection only.
How to trigger NVIDIA Smooth Motion works through SVP? Thats new AI thing that working with optical flow, according to NVIDIA ofcourse!
no idea, there's no API for that yet
@Chainik, is SVP updated with tensort that is compatible with rtx 5XXX ?
@Chainik, is SVP updated with tensort that is compatible with rtx 5XXX ?
Actually not, but 15.8 supports 50 series!
Managed to get a 5090 day1, would be really happy to be a betatester to all your experiments XD
Coming from a 1070ti and honestly can't wait to test properly RIFE (tested it with a 4060 max and the results were amazing)
It should come this week (or next one). Any suggestion?
Managed to get a 5090 day1, would be really happy to be a betatester to all your experiments XD
Coming from a 1070ti and honestly can't wait to test properly RIFE (tested it with a 4060 max and the results were amazing)It should come this week (or next one). Any suggestion?
https://github.com/AmusementClub/vs-mlrt/releases/
Also dont forget update to v15.8: Blackwell support, latest TensorRT!
Is there a model that can do 2x with a 1440p60 video on a 4080S? I know I can downsize with SVP but I'd like to avoid doing that.
Is there a model that can do 2x with a 1440p60 video on a 4080S? I know I can downsize with SVP but I'd like to avoid doing that.
The 4.4v2 model best on performance, but overall, quality remains questionable! Try everything from 4.4v2 to 4.9v2 or switch to 4.15v2 lite (4.9 perf)!
Just to inform you that vs-rife has got major performance improvement in recent releases. Tested with the latest vs-mlrt (v15.8) and vs-rife (v5.6.0) on my low-end i5-13400F and 4060 Ti:
# from vsmlrt import BackendV2, RIFE, RIFEModel
# clip = RIFE(clip, model=RIFEModel.v4_25, backend=BackendV2.TRT(num_streams=3, fp16=True, output_format=1), _implementation=2)
1080p: 84.23 fps
720p: 210.37 fps# from vsrife import rife
# clip = rife(clip, model="4.25", trt=True)
1080p: 102.99 fps
720p: 249.36 fpsJust to inform you that vs-rife has got major performance improvement in recent releases. Tested with the latest vs-mlrt (v15.8) and vs-rife (v5.6.0) on my low-end i5-13400F and 4060 Ti
Hmm... is that vs-rife (v5.6.0) something to update in SVP?
Today I received my RTX 5080 and SVP has stopped working, while with my previous RTX 4070 it works perfectly. I have seen that some components need to be updated, but the application does not show any updates. Will SVP be updated? Can anyone describe step-by-step how to do it manually if necessary? Sorry, I don't speak English, I'm using Copilot to translate.
Today I received my RTX 5080 and SVP has stopped working, while with my previous RTX 4070 it works perfectly. I have seen that some components need to be updated, but the application does not show any updates. Will SVP be updated? Can anyone describe step-by-step how to do it manually if necessary? Sorry, I don't speak English, I'm using Copilot to translate.
1) Download this
2) Unpack only vsmlrt-cuda folder from archive to ...\SVP 4\rife
2) Unpack only vsmlrt-cuda folder from archive to ...\SVP 4\rife
I also see that the RIFE developer added some extra parameters for the new TensorRT version, not sure if mandatory:
https://gyazo.com/87bdd69a3c5e6d3f934b003cbe4a71b1
> Hmm... is that vs-rife (v5.6.0) something to update in SVP?
nope
PyTorch dependency is not what we want
let's wait for vsmlrt to catch up...
The 4.4v2 model best on performance, but overall, quality remains questionable! Try everything from 4.4v2 to 4.9v2 or switch to 4.15v2 lite (4.9 perf)!
Tried all of them but it seems current RIFE that comes with SVP isn't quite there to handle it. Both 4.15l and 4.16l fall just 15 fps short of hitting the 120 target.
Now I'm eagerly awaiting for the update as that looks like just what's needed to get there.
RickyAstle98 wrote:The 4.4v2 model best on performance, but overall, quality remains questionable! Try everything from 4.4v2 to 4.9v2 or switch to 4.15v2 lite (4.9 perf)!
Tried all of them but it seems current RIFE that comes with SVP isn't quite there to handle it. Both 4.15l and 4.16l fall just 15 fps short of hitting the 120 target.
Now I'm eagerly awaiting for the update as that looks like just what's needed to get there.
This is strange, my 4070 handles 4.15l v2 1440p 60x2 without issues...
This is strange, my 4070 handles 4.15l v2 1440p 60x2 without issues...
I'm probably doing something wrong because I can't get v2 model (downloaded here) to run at all:
Last bits of trtexec_250207_183007.log:
[02/07/2025-18:30:10] [W] [TRT] Could not read timing cache from: C:\Users\...\AppData\Roaming\SVP4\cache\Program Files (x86)/SVP 4/rife\models\rife\rife_v4.15_lite.onnx.2560x1440_fp16_no-tf32_trt-100400_I-fp16_O-fp16_NVIDIA-GeForce-RTX-4080-SUPER_b38f6198.engine.cache. A new timing cache will be generated and written.
[02/07/2025-18:30:10] [E] Error[2]: [castLayer.cpp::nvinfer1::CastLayer::validateTypes::45] Error Code 2: Internal Error (Assertion !mOutputTypes.at(0).hasValue() || mOutputTypes.at(0).value() == params.toType failed. )
[02/07/2025-18:30:10] [E] Engine could not be created from network
[02/07/2025-18:30:10] [E] Building engine failed
[02/07/2025-18:30:10] [E] Failed to create engine from model or file.
[02/07/2025-18:30:10] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v100500] [b18] # C:/Program Files (x86)/SVP 4/rife\vsmlrt-cuda\trtexec --onnx=C:/Program Files (x86)/SVP 4/rife\models\rife\rife_v4.15_lite.onnx --timingCacheFile=C:\Users\...\AppData\Roaming\SVP4\cache\Program Files (x86)/SVP 4/rife\models\rife\rife_v4.15_lite.onnx.2560x1440_fp16_no-tf32_trt-100400_I-fp16_O-fp16_NVIDIA-GeForce-RTX-4080-SUPER_b38f6198.engine.cache --device=0 --saveEngine=C:\Users\...\AppData\Roaming\SVP4\cache\Program Files (x86)/SVP 4/rife\models\rife\rife_v4.15_lite.onnx.2560x1440_fp16_no-tf32_trt-100400_I-fp16_O-fp16_NVIDIA-GeForce-RTX-4080-SUPER_b38f6198.engine --shapes=input:1x11x1440x2560 --fp16 --tacticSources=-CUBLAS,-CUBLAS_LT,-CUDNN,+EDGE_MASK_CONVOLUTIONS,+JIT_CONVOLUTIONS --useCudaGraph --noDataTransfers --noTF32 --inputIOFormats=fp16:chw --outputIOFormats=fp16:chw --layerPrecisions=*:fp16 --layerOutputTypes=*:fp16 --precisionConstraints=obey --builderOptimizationLevel=3TRT 10.8 libraries available via SVP update
RickyAstle98 wrote:This is strange, my 4070 handles 4.15l v2 1440p 60x2 without issues...
I'm probably doing something wrong because I can't get v2 model (downloaded here) to run at all:
v2 models should be put in rife_v2 folder
TRT 10.8 libraries available via SVP update
Thanks.
On RTX 4070, it broke rife 4.25 heavy models (also v2).
4.25, 4.25lite and v2 are working, 4.26 and 4.26 heavy are ok.
Resolution 1280x768 or 1920x1080
I would wait if others have the same issue before doing anythings else.
[02/07/2025-18:54:55] [W] [TRT] Could not read timing cache from: C:/Prog/SVP 4/rife\models\rife\rife_v4.25_heavy.onnx.1280x768_fp16_no-tf32_trt-100800_I-fp16_O-fp16_NVIDIA-GeForce-RTX-4070-SUPER_733fcbe3.engine.cache. A new timing cache will be generated and written.
[02/07/2025-18:54:55] [E] Error[2]: [castLayer.cpp::nvinfer1::CastLayer::validateTypes::45] Error Code 2: Internal Error (Assertion !mOutputTypes.at(0).hasValue() || mOutputTypes.at(0).value() == params.toType failed. )
[02/07/2025-18:54:55] [E] Engine could not be created from network
[02/07/2025-18:54:55] [E] Building engine failed
[02/07/2025-18:54:55] [E] Failed to create engine from model or file.
[02/07/2025-18:54:55] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v100800] [b43] # C:/Prog/SVP 4/rife\vsmlrt-cuda\trtexec --onnx=C:/Prog/SVP 4/rife\models\rife\rife_v4.25_heavy.onnx --timingCacheFile=C:/Prog/SVP 4/rife\models\rife\rife_v4.25_heavy.onnx.1280x768_fp16_no-tf32_trt-100800_I-fp16_O-fp16_NVIDIA-GeForce-RTX-4070-SUPER_733fcbe3.engine.cache --device=0 --saveEngine=C:/Prog/SVP 4/rife\models\rife\rife_v4.25_heavy.onnx.1280x768_fp16_no-tf32_trt-100800_I-fp16_O-fp16_NVIDIA-GeForce-RTX-4070-SUPER_733fcbe3.engine --shapes=input:1x11x768x1280 --fp16 --tacticSources=-CUBLAS,-CUBLAS_LT,-CUDNN,+EDGE_MASK_CONVOLUTIONS,+JIT_CONVOLUTIONS --useCudaGraph --noDataTransfers --noTF32 --inputIOFormats=fp16:chw --outputIOFormats=fp16:chw --layerPrecisions=*:fp16 --layerOutputTypes=*:fp16 --precisionConstraints=obey --builderOptimizationLevel=3> it broke rife 4.25 heavy models
yeah, doesn't work for me too
> it broke rife 4.25 heavy models
yeah, doesn't work for me too
I have 4070 too, just rename 4.25 heavy model to 4.30 and everything works!
SmoothVideo Project → Using SVP → New RIFE filter - 3x faster AI interpolation possible in SVP!!!
Powered by PunBB, supported by Informer Technologies, Inc.