Hybrid(Windows/Linux/Mac): Input -> x264/x265/Xvid/VP8/VP9

[Selur]

Yes, both will save the engine files in the same folder. I know they name them differently. But I never really looked at how they are named.

[Platos]

Ok, then i could just filter by name and delete the torch-one (probably there are no .engine files from torch, because it crashed before).

Thanks

[Selur]

Probably.

[Platos]

How can i passthrough (all) subtitles (in correct order) like i can do with audio? So how can i make sure, that all subtitles stay after encoding ? I don't get it. Does someone know that ?

[Selur]

see "Config->Automation->Subtitle"

[Platos]

Thanks !

[Selur]

Also read: https://forum.selur.net/thread-2473.html

[Platos]

Question: I saw my default Temp Path ist \AppData\Local\Temp
So that means it's no own folder. Do you know if there is any way to delete Temp-Files of Hybrid, if it is saved there?
And you mentioned above some weeks ago i could delete .engine files from time to time. But where is the folder? I did not find it. I saw some .egine files in my local\temp folder, but they are named not like i remember in the past (in past they were named with my Graphicscard name)

[Selur]

Config->Paths allows setting input, output, temp, engine and debut output path

Hybrid itself has nothing to do with the naming of the .engine files, those are created by the filters. Some filters but the graphic card name in that name others don't.

Cu Selur

[Platos]

Thanks!

Question: I remember you had RTX4080 and 16-core AMD CPU: Do you have also problem with GPU Usage on mlrt RIFE and torchRIFe? I have a 13900k and RTX4080 and now RTX5080. With both i have low GPU Usage. CPU is ~ around 75-80%, but not 100% (when i had a intel i5 12400f and a RTX4080 my cpu was 100% completly). But GPU Usage ist around 65% only with 2 or 3 Streams (doesnt matter).

So do you have same behavior?

BTW: I noticed a crash on mlrtRIFe without FP16 with TensorRT (without RTX). Vapoursynth preview says:

VapourSynth plugins manager: Failed to load vapoursynth library!
Please set up the library search paths in settings.
Failed to evaluate the script:
Python exception: operator (): engine deserialization failed

Traceback (most recent call last):
File "src/cython/vapoursynth.pyx", line 3378, in vapoursynth._vpy_evaluate
File "src/cython/vapoursynth.pyx", line 3379, in vapoursynth._vpy_evaluate
File "C:\Users\Gaming-Tower\AppData\Local\Temp\Hybrid\tempPreviewVapoursynthFile23_59_58_293.vpy", line 49, in 
clip = vsmlrt.RIFE(clip, multi=Fraction(2.5025), model=425, backend=Backend.TRT(fp16=False,device_id=0,bf16=False,verbose=True,use_cuda_graph=False, num_streams=3,builder_optimization_level=3,engine_folder="C:/Users/Gaming-Tower/Documents/Hybrid/Engine-Files")) # new fps: 60
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:/Program Files/Hybrid/64bit/vs-mlrt/vsmlrt.py", line 1410, in RIFE
output0 = RIFEMerge(
^^^^^^^^^^
File "C:/Program Files/Hybrid/64bit/vs-mlrt/vsmlrt.py", line 1144, in RIFEMerge
return inference_with_fallback(
^^^^^^^^^^^^^^^^^^^^^^^^
File "C:/Program Files/Hybrid/64bit/vs-mlrt/vsmlrt.py", line 2979, in inference_with_fallback
raise e
File "C:/Program Files/Hybrid/64bit/vs-mlrt/vsmlrt.py", line 2956, in inference_with_fallback
ret = _inference(
^^^^^^^^^^^
File "C:/Program Files/Hybrid/64bit/vs-mlrt/vsmlrt.py", line 2819, in _inference
ret = core.trt.Model(
^^^^^^^^^^^^^^^
File "src/cython/vapoursynth.pyx", line 3114, in vapoursynth.Function.__call__
vapoursynth.Error: operator (): engine deserialization failed

And vapoursynth script view says:

# Imports
import vapoursynth as vs
# getting Vapoursynth core
import logging
import site
from fractions import Fraction
import sys
import os
core = vs.core
# Import scripts folder
scriptPath = 'C:/Program Files/Hybrid/64bit/vsscripts'
sys.path.insert(0, os.path.abspath(scriptPath))
os.environ["CUDA_MODULE_LOADING"] = "LAZY"
# Force logging to std:err
logging.StreamHandler(sys.stderr)
# loading plugins
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vs-mlrt/vstrt.dll")
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/Support/akarin.dll")
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/MiscFilter/MiscFilters/MiscFilters.dll")
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/SourceFilter/LSmashSource/LSMASHSource.dll")
# Import scripts
from importlib.machinery import SourceFileLoader
vsmlrt = SourceFileLoader('vsmlrt', 'C:/Program Files/Hybrid/64bit/vs-mlrt/vsmlrt.py').load_module()
from vsmlrt import Backend
import validate
# Source: 'C:\Users\Gaming-Tower\Downloads\test1.mkv'
# Current color space: YUV420P10, bit depth: 10, resolution: 3840x1600, frame rate: 23.976fps, scanorder: progressive, yuv luminance scale: limited, matrix: 2020ncl, transfer: smpte2084, primaries: bt.2020, format: HEVC
# Loading C:\Users\Gaming-Tower\Downloads\test1.mkv using LWLibavSource
clip = core.lsmas.LWLibavSource(source="C:/Users/Gaming-Tower/Downloads/test1.mkv", format="YUV420P10", stream_index=0, cache=0, fpsnum=24000, fpsden=1001, prefer_hw=0)
frame = clip.get_frame(0)
# setting color matrix to 2020ncl.
clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_BT2020_NCL)
# setting color transfer (vs.TRANSFER_ST2084), if it is not set.
if validate.transferIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Transfer=vs.TRANSFER_ST2084)
# setting color primaries info (to vs.PRIMARIES_BT2020), if it is not set.
if validate.primariesIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT2020)
# setting color range to TV (limited) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED)
# making sure frame rate is set to 23.976fps
clip = core.std.AssumeFPS(clip=clip, fpsnum=24000, fpsden=1001)
# making sure the detected scan type is set (detected: progressive)
clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive
clip = core.misc.SCDetect(clip=clip,threshold=0.070)
# adjusting color space from YUV420P10 to RGBS for vsRIFEmlrt
clip = core.resize.Bicubic(clip=clip, format=vs.RGBS, matrix_in_s="2020ncl", range_s="limited")
# adjusting frame count&rate with RIFE (mlrt)
clip = vsmlrt.RIFE(clip, multi=Fraction(2.5025), model=425, backend=Backend.TRT(fp16=False,device_id=0,bf16=False,verbose=True,use_cuda_graph=False, num_streams=3,builder_optimization_level=3,engine_folder="C:/Users/Gaming-Tower/Documents/Hybrid/Engine-Files")) # new fps: 60
# adjusting output color from: RGBS to YUV420P10 for x265Model
clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P10, matrix_s="2020ncl", range_s="limited", dither_type="error_diffusion")
# set output frame rate to 60fps (progressive)
clip = core.std.AssumeFPS(clip=clip, fpsnum=60, fpsden=1)
# output
clip.set_output()

And i did also encounter an error with torchRIFE (doesnt matter if FPS 16 or not or if scale 0.5 or not or if TensorRT or not):

2025-07-12 00:10:34.442
VapourSynth plugins manager: Failed to load vapoursynth library!
Please set up the library search paths in settings.
2025-07-12 00:10:36.331
C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\torch\cuda\__init__.py:287: UserWarning:
NVIDIA GeForce RTX 5080 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.
If you want to use the NVIDIA GeForce RTX 5080 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

warnings.warn(

C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\torch\cuda\__init__.py:287: UserWarning:
NVIDIA GeForce RTX 5080 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.
If you want to use the NVIDIA GeForce RTX 5080 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

warnings.warn(

2025-07-12 00:10:36.407
Failed to evaluate the script:
Python exception: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.


Traceback (most recent call last):
File "src/cython/vapoursynth.pyx", line 3378, in vapoursynth._vpy_evaluate
File "src/cython/vapoursynth.pyx", line 3379, in vapoursynth._vpy_evaluate
File "C:\Users\Gaming-Tower\AppData\Local\Temp\Hybrid\tempPreviewVapours ynthFile00_10_34_384.vpy", line 41, in
clip = rife(clip, fps_num=60, fps_den=1, scale=0.5, sc=True, sc_threshold=0.070) # new fps: 60
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^
File "contextlib.py", line 81, in inner
File "C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\vsrife\__init__.py", line 633, in rife
tenHorizontal = torch.linspace(-1.0, 1.0, pw, dtype=torch.float, device=device)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^
RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

And the script view says:

# Imports
import vapoursynth as vs
# getting Vapoursynth core
import logging
import sys
import os
core = vs.core
# Import scripts folder
scriptPath = 'C:/Program Files/Hybrid/64bit/vsscripts'
sys.path.insert(0, os.path.abspath(scriptPath))
# Force logging to std:err
logging.StreamHandler(sys.stderr)
# loading plugins
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/MiscFilter/MiscFilters/MiscFilters.dll")
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/SourceFilter/LSmashSource/LSMASHSource.dll")
# Import scripts
import validate
# Source: 'C:\Users\Gaming-Tower\Downloads\test1.mkv'
# Current color space: YUV420P10, bit depth: 10, resolution: 3840x1600, frame rate: 23.976fps, scanorder: progressive, yuv luminance scale: limited, matrix: 2020ncl, transfer: smpte2084, primaries: bt.2020, format: HEVC
# Loading C:\Users\Gaming-Tower\Downloads\test1.mkv using LWLibavSource
clip = core.lsmas.LWLibavSource(source="C:/Users/Gaming-Tower/Downloads/test1.mkv", format="YUV420P10", stream_index=0, cache=0, fpsnum=24000, fpsden=1001, prefer_hw=0)
frame = clip.get_frame(0)
# setting color matrix to 2020ncl.
clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_BT2020_NCL)
# setting color transfer (vs.TRANSFER_ST2084), if it is not set.
if validate.transferIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Transfer=vs.TRANSFER_ST2084)
# setting color primaries info (to vs.PRIMARIES_BT2020), if it is not set.
if validate.primariesIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT2020)
# setting color range to TV (limited) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED)
# making sure frame rate is set to 23.976fps
clip = core.std.AssumeFPS(clip=clip, fpsnum=24000, fpsden=1001)
# making sure the detected scan type is set (detected: progressive)
clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive
from vsrife import rife
# adjusting color space from YUV420P10 to RGBS for vsTorchRIFE
clip = core.resize.Bicubic(clip=clip, format=vs.RGBS, matrix_in_s="2020ncl", range_s="limited")
# adjusting frame count&rate with RIFE (torch)
clip = rife(clip, fps_num=60, fps_den=1, scale=0.5, sc=True, sc_threshold=0.070) # new fps: 60
# adjusting output color from: RGBS to YUV420P10 for x265Model
clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P10, matrix_s="2020ncl", range_s="limited", dither_type="error_diffusion")
# set output frame rate to 60fps (progressive)
clip = core.std.AssumeFPS(clip=clip, fpsnum=60, fpsden=1)
# output
clip.set_output()

Do you need .debug file too ? I have hybrid 2025.06.29.1 and newest addons from t hat time.

[Selur]

I always need a debug output for proper details.

NVIDIA GeForce RTX 5080 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.

try latest dev, don't have a 5xxx card, but I know older torch and vs-mlrt do not support them.
Also:
a. Try clearing your engine files and temp folders.
b. " Python exception: CUDA error: no kernel image is available for execution on the device" does sound like a driver issue. Try 'studio drivers'.

Cu Selur

[Platos]

vsmlrt: deleting .engine files did solve the problem. Thank you!

torchrife: I installed newest studiodriver (have always studio) and newest dev version and i had already newest vapoursynth and vsmlrt and so on. Did create new temp-folder (so the files from APPDATA should be no problem). It did not work with default TorchRIFE-settings (or any other). Same error with rtx5080 does not work.

debug-file is uploaded.

Are the Vapoursynth and vsmlrt the newest on your google drive or is there any newer i can download from github or so? Because the files here looks not the same (not same file size) so i think i can not just download them: https://github.com/AmusementClub/vs-mlrt/releases
Could you tell me, which file i could download for trying out ?

Or do i just have to wait for them to update their vsmlrt/vapoursynth ?

these is the error from vapoursynth preview:

2025-07-12 10:52:18.363
VapourSynth plugins manager: Failed to load vapoursynth library!
Please set up the library search paths in settings.
2025-07-12 10:52:20.310
C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\torch\cuda\__init__.py:287: UserWarning: 
NVIDIA GeForce RTX 5080 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.
If you want to use the NVIDIA GeForce RTX 5080 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

warnings.warn(

C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\torch\cuda\__init__.py:287: UserWarning: 
NVIDIA GeForce RTX 5080 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.
If you want to use the NVIDIA GeForce RTX 5080 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

warnings.warn(

2025-07-12 10:52:20.411
Failed to evaluate the script:
Python exception: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.


Traceback (most recent call last):
File "src/cython/vapoursynth.pyx", line 3378, in vapoursynth._vpy_evaluate
File "src/cython/vapoursynth.pyx", line 3379, in vapoursynth._vpy_evaluate
File "C:\Users\Gaming-Tower\AppData\Local\Temp\Hybrid\tempPreviewVapoursynthFile10_52_18_302.vpy", line 41, in 
clip = rife(clip, sc=True, sc_threshold=0.100) # new fps: 47.952
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "contextlib.py", line 81, in inner
File "C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Program Files\Hybrid\64bit\Vapoursynth\Lib\site-packages\vsrife\__init__.py", line 633, in rife
tenHorizontal = torch.linspace(-1.0, 1.0, pw, dtype=torch.float, device=device)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

And these is the code from vapoursynth script-view:

# Imports
import vapoursynth as vs
# getting Vapoursynth core
import logging
import sys
import os
core = vs.core
# Import scripts folder
scriptPath = 'C:/Program Files/Hybrid/64bit/vsscripts'
sys.path.insert(0, os.path.abspath(scriptPath))
# Force logging to std:err
logging.StreamHandler(sys.stderr)
# loading plugins
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/MiscFilter/MiscFilters/MiscFilters.dll")
core.std.LoadPlugin(path="C:/Program Files/Hybrid/64bit/vsfilters/SourceFilter/LSmashSource/LSMASHSource.dll")
# Import scripts
import validate
# Source: 'C:\Users\Gaming-Tower\Downloads\$Filme-Serien-Animes&Dokus$\[Downloadordner]\test1.mkv'
# Current color space: YUV420P10, bit depth: 10, resolution: 3840x1600, frame rate: 23.976fps, scanorder: progressive, yuv luminance scale: limited, matrix: 2020ncl, transfer: smpte2084, primaries: bt.2020, format: HEVC
# Loading C:\Users\Gaming-Tower\Downloads\$Filme-Serien-Animes&Dokus$\[Downloadordner]\test1.mkv using LWLibavSource
clip = core.lsmas.LWLibavSource(source="C:/Users/Gaming-Tower/Downloads/$Filme-Serien-Animes&Dokus$/[Downloadordner]/test1.mkv", format="YUV420P10", stream_index=0, cache=0, fpsnum=24000, fpsden=1001, prefer_hw=0)
frame = clip.get_frame(0)
# setting color matrix to 2020ncl.
clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_BT2020_NCL)
# setting color transfer (vs.TRANSFER_ST2084), if it is not set.
if validate.transferIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Transfer=vs.TRANSFER_ST2084)
# setting color primaries info (to vs.PRIMARIES_BT2020), if it is not set.
if validate.primariesIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT2020)
# setting color range to TV (limited) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED)
# making sure frame rate is set to 23.976fps
clip = core.std.AssumeFPS(clip=clip, fpsnum=24000, fpsden=1001)
# making sure the detected scan type is set (detected: progressive)
clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive
from vsrife import rife
# adjusting color space from YUV420P10 to RGBS for vsTorchRIFE
clip = core.resize.Bicubic(clip=clip, format=vs.RGBS, matrix_in_s="2020ncl", range_s="limited")
# adjusting frame count&rate with RIFE (torch)
clip = rife(clip, sc=True, sc_threshold=0.100) # new fps: 47.952
# adjusting output color from: RGBS to YUV420P10 for x265Model
clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P10, matrix_s="2020ncl", range_s="limited", dither_type="error_diffusion")
# set output frame rate to 47.952fps (progressive)
clip = core.std.AssumeFPS(clip=clip, fpsnum=48000, fpsden=1001)
# output
clip.set_output()

[Selur]

Side note: I strongly advise against using %APPDATA% as temp folder, since virus scanners tend to be overzealous in those folders.
"so the files from APPDATA should be no problem" only if the used filter does not check those folders,... (haven't checked the source code)

Are the Vapoursynth and vsmlrt the newest on your google drive or is there any newer i can download from github or so? Because the files here looks not the same (not same file size) so i think i can not just download them: https://github.com/AmusementClub/vs-mlrt/releases
Could you tell me, which file i could download for trying out ?

vs-mlrt_2025.06.15 should contain the files from https://github.com/AmusementClub/vs-mlrt/releases/tag/v15.12 + https://github.com/AmusementClub/vs-mlrt/releases/tag/external-models
If something is different, it's likely a mistake on my part.

Latest vs-mlrt source is usually newer.
With my GeForce RTX 4080 (576.80 Studio Drivers) both vs-mlrt and torch filters work fine.
(I'm not planning on buying a 5080 since I already paid enough for my 4080 )

For general clarification, about what "Frame->Interpolation->Frame Interpolation" uses:

• RIFE <> https://github.com/styler00dollar/VapourSynth-RIFE-ncnn-Vulkan
• TorchRIFE <> https://github.com/HolyWu/vs-rife
• RIFEmlrt <> https://github.com/AmusementClub/vs-mlrt

CUDA error: no kernel image is available for execution on the device

usually hints to a problem between CUDA enviroment and drivers.
Make sure to delete the old Vapoursynth folder before using the one from the torch-addon (like the README states), otherwise the environment can be messed up.
Same for vs-mlrt delete old vs-mlrt folder.
Extracting the extension into the Hybrid folder and simply overwriting duplicate files likely will cause issues.

Cu Selur

[Platos]

Ok, i did complete deinstall Hybrid and all Files of it and reinstall it and after installing and starting (and closing) i did delete vapoursynth, then extract the addons somewhere else and then copy (so move) the folder into the 64bit folder. the same for vsmlrt, but because of reinstallation it was not there, so no deleting before. Onnx and vsgan models i did also extract first and then move the folder into hybrid.

Problem with torchRIFE is the same. (RTX 5080...). On TorchRIFE website you linked, i don't understand, what i have to download and where in hybrid to copy. i think i will just wait, till newer versions comes out and then download it from you
because "pip install -U vsrife" sounds like a global installation when using phyton stuff. I don't know how i can download that as a folder/into a folder ^^

[Selur]

If you want to manually change stuff in the torch setup:
1. open a Terminal in Hybrid/64bit/Vapoursynth (there a portable Python version is installed)
2. instead of "pip install -U vsrife" call "python -m pip install -U vsrife"

If you want to manually change stuff in the vs-mlrt setup:
a. if you use the latest git you need to compile stuff yourself
b. if you just want to use the latest vsmlrt.py you can download the one from the git and replace the existing one.
In general the vs-mlrt addon (ignoring the onnx_model folder) in Hybrid simply is a 'vs-mlrt'-folder into which all files (that are not sources) from the lastest release are extracted.

Cu Selur

ps.: Manually setting up the torch addon is also possible, and basically follows https://pastebin.com/1fv6efmQ (with updated files; didn't write down a new version last time I updated the addon)

[Platos]

Ok thanks. If i have time next days, i will try it. If not, i surely can need this guide maybe in future, thank you!