Deoldify Vapoursynth filter

[Selur]

Hybrid_dev_2026.09.26-162846 und VapoursynthR79_torch_2026.09.29 sind oben und sollten die entsprechenden Dateien enthalten.

Cu Selur

[Dan64]

Tested  Hybrid 2026.09.26.1 with Vapoursynth R79 and it works

Thanks,
Dan

[Selur]

Thanks for the feedback and much fun.

Cu Selur

Ps.: since R80 drops API-3 support, Hybrid atm. will stick to R79 (got ~40 filters in Hybrid which are still API-3 only)

[hallomanbh]

2026-09-30 23:33:52.061
Script was successfully evaluated. Output video info:
Frames: 265 | Time: 0:00:08.842 | Size: 2560x720 | FPS: 30000/1001 = 29.97 | Format: YUV420P10
2026-09-30 23:34:26.535
Failed to evaluate the script:
Python exception: Numba needs NumPy 2.3 or less. Got NumPy 2.5.

Traceback (most recent call last):
File "vapoursynth.pyx", line 3832, in vapoursynth._vpy_evaluate
File "vapoursynth.pyx", line 3833, in vapoursynth._vpy_evaluate
File "C:\Users\halloman\AppData\Local\Temp\tempPreviewVapoursynthFile23_33_55_491.vpy", line 15, in
import vshavc as havc
File "c:\program files\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\__init__.py", line 41, in
from vshavc.vsslib.vsmodels import vs_sc_deoldify, vs_sc_ddcolor, vs_colormnet, vs_deepex, vs_deepremaster
File "c:\program files\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\vsslib\vsmodels.py", line 27, in
from vshavc.deepex import deepex_colorizer, ModelColorizer
File "c:\program files\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\deepex\__init__.py", line 31, in
from vshavc.deepex.utils.util_distortion import CenterPad, Normalize, RGB2Lab, ToTensor
File "c:\program files\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\deepex\utils\util_distortion.py", line 15, in
from numba import jit, u1, u2
File "c:\program files\hybrid\64bit\vapoursynth\Lib\site-packages\numba\__init__.py", line 59, in
_ensure_critical_deps()
File "c:\program files\hybrid\64bit\vapoursynth\Lib\site-packages\numba\__init__.py", line 45, in _ensure_critical_deps
raise ImportError(msg)
ImportError: Numba needs NumPy 2.3 or less. Got NumPy 2.5.

[Selur]

Something is wrong there.
I just checked, Hybrids torch-add-on comes with:

numba 0.63.1 numexpr 2.14.1 numpy 2.2.6

=> no clue, where NumPy 2.5 comes from on your system.

Ahh,.. I got it, Hybrids dev does come with numpy 2.5.2 and you didn't properly install the torch add-on.

Installation of 'torch'-addon

Delete old Hybrid/64bit/Vapoursynth-folder.
Extract the torch-addon into your Hybrid/64bit-folder.
Extract the vsgan_models into your Hybrid/64bit-folder. (belongs to vsgan which comes with the torch addon)

from the README.md

[hallomanbh]

yes updated working

[Selur]

@Dan64: could you provide a Python 3.14 compatile spatial_correlation_sampler wheel?

Cu Selur

[Dan64]

@Dan64: could you provide a Python 3.14 compatile spatial_correlation_sampler wheel?

Cu Selur

For which version of torch ? please provide the pip installation command, so I'm sure to build against the same torch version that you are using.

Thanks,
Dan

[Selur]

Here's my pip list: https://pastebin.com/CFU9NYLX
I use Python 3.14.8 and the torch stack (CUDA 13) was installed with:

python.exe -m pip install -U packaging setuptools wheel python.exe -m pip install torch==2.14.1 torchvision==0.29.1 --index-url https://download.pytorch.org/whl/cu130 python.exe -m pip install torch_tensorrt==2.14.0 --extra-index-url https://pypi.nvidia.com python.exe -m pip install "cupy-cuda13x[ctk]"

`torch_tensorrt` 2.14.0 (PyPI, cp314) is the cu130 build and pulls `tensorrt-cu13` 11.1.0.106.
`cupy-cuda13x[ctk]` adds the CUDA 13.4 toolkit wheels (`cuda-toolkit`, `nvidia-cuda-runtime`, …) with the headers;

Cu Selur

[Dan64]

Here's my pip list: https://pastebin.com/CFU9NYLX
I use Python 3.14.8 and the torch stack (CUDA 13) was installed with:

python.exe -m pip install -U packaging setuptools wheel python.exe -m pip install torch==2.14.1 torchvision==0.29.1 --index-url https://download.pytorch.org/whl/cu130 python.exe -m pip install torch_tensorrt==2.14.0 --extra-index-url https://pypi.nvidia.com python.exe -m pip install "cupy-cuda13x[ctk]"

`torch_tensorrt` 2.14.0 (PyPI, cp314) is the cu130 build and pulls `tensorrt-cu13` 11.1.0.106.
`cupy-cuda13x[ctk]` adds the CUDA 13.4 toolkit wheels (`cuda-toolkit`, `nvidia-cuda-runtime`, …) with the headers;

Cu Selur

I attached the build compiled for the CUDA 13.0

Dan