v0.4.0

What's New

ccrestoration and ccvfi are no longer maintained, consider using cccv instead.
In cccv, we have implemented most of the features from ccrestoration and ccvfi, Enjoy!

Base Environment

• system: Ubuntu 22.04
• GCC/G++: 13
• python: 3.10
• FFmpeg: 8.0
• Static FFmpeg: 8.0 (/static-ffmpeg/ffmpeg)
• VapourSynth: R70
• x264: latest
• x265: 4.1
• svt-av1-psy: latest
• aom: latest
• libvpx: latest
• fdk-aac: latest
• libass: latest

AI Environment

• CUDA toolkit: 12.9
• PyTorch: 2.8.0+cu128 ['sm_70', 'sm_75', 'sm_80', 'sm_86', 'sm_90', 'sm_100', 'sm_120']
• cupy: cuda12x

VapourSynth Python Plugin List

mbfunc
cccv
vsutil
mvsfunc

VapourSynth C++ Plugin List

bestsource.so
libaddgrain.so
libakarin.so
libassrender.so
libbilateral.so
libbm3dcpu.so
libbm3dcuda.so
libbm3dcuda_rtc.so
libboxblur.so
libbwdif.so
libcas.so
libctmf.so
#libd2vsource.so
libdctfilter.so
libdescale.so
libdfttest2_cpu.so
libdfttest2_cuda.so
libdfttest2_nvrtc.so
libeedi2.so
libeedi3m.so
libffms2.so
libfillborders.so
libfluxsmooth.so
libfmtconv.so
libhqdn3d.so
libils.so
libmiscfilters.so
libmvtools.so
libneo-dfttest.so
libneo-f3kdb.so
libneo-fft3d.so
libnnedi3.so
libremapframes.so
libremovegrain.so
libretinex.so
libsangnom.so
libtcanny.so
libtedgemask.so
libttempsmooth.so
libvsnlm_cuda.so
libvsznedi3.so

v0.3.2

• build: bump PyTorch to 2.8.0+cu128 ['sm_70', 'sm_75', 'sm_80', 'sm_86', 'sm_90', 'sm_100', 'sm_120']
• build: downgrade nv-codec-headers to 12.2 (driver 550+ support)

Base Environment

• system: Ubuntu 22.04
• GCC/G++: 13
• python: 3.10
• FFmpeg: 8.0
• Static FFmpeg: 8.0 (/static-ffmpeg/ffmpeg)
• VapourSynth: R70
• x264: latest
• x265: 4.1
• svt-av1-psy: latest
• aom: latest
• libvpx: latest
• fdk-aac: latest
• libass: latest

AI Environment

• CUDA toolkit: 12.9
• PyTorch: 2.8.0+cu128 ['sm_70', 'sm_75', 'sm_80', 'sm_86', 'sm_90', 'sm_100', 'sm_120']
• cupy: cuda12x

VapourSynth Python Plugin List

mbfunc
ccrestoration
ccvfi
vsutil
mvsfunc

VapourSynth C++ Plugin List

bestsource.so
libaddgrain.so
libakarin.so
libassrender.so
libbilateral.so
libbm3dcpu.so
libbm3dcuda.so
libbm3dcuda_rtc.so
libboxblur.so
libbwdif.so
libcas.so
libctmf.so
#libd2vsource.so
libdctfilter.so
libdescale.so
libdfttest2_cpu.so
libdfttest2_cuda.so
libdfttest2_nvrtc.so
libeedi2.so
libeedi3m.so
libffms2.so
libfillborders.so
libfluxsmooth.so
libfmtconv.so
libhqdn3d.so
libils.so
libmiscfilters.so
libmvtools.so
libneo-dfttest.so
libneo-f3kdb.so
libneo-fft3d.so
libnnedi3.so
libremapframes.so
libremovegrain.so
libretinex.so
libsangnom.so
libtcanny.so
libtedgemask.so
libttempsmooth.so
libvsnlm_cuda.so
libvsznedi3.so

v0.3.1

• Add Static FFmpeg in /static-ffmpeg/ffmpeg
• Support hevc_nvenc (NVIDIA GPU video encoding)
• Fix docker build

Base Environment

• system: Ubuntu 22.04
• GCC/G++: 13
• python: 3.10
• FFmpeg: 8.0
• Static FFmpeg: 8.0 (/static-ffmpeg/ffmpeg)
• VapourSynth: R70
• x264: latest
• x265: 4.1
• svt-av1-psy: latest
• aom: latest
• libvpx: latest
• fdk-aac: latest
• libass: latest

AI Environment

• CUDA toolkit: 12.9
• PyTorch: 2.7.1+cu128
• cupy: cuda12x

VapourSynth Python Plugin List

mbfunc
ccrestoration
ccvfi
vsutil
mvsfunc

VapourSynth C++ Plugin List

bestsource.so
libaddgrain.so
libakarin.so
libassrender.so
libbilateral.so
libbm3dcpu.so
libbm3dcuda.so
libbm3dcuda_rtc.so
libboxblur.so
libbwdif.so
libcas.so
libctmf.so
#libd2vsource.so
libdctfilter.so
libdescale.so
libdfttest2_cpu.so
libdfttest2_cuda.so
libdfttest2_nvrtc.so
libeedi2.so
libeedi3m.so
libffms2.so
libfillborders.so
libfluxsmooth.so
libfmtconv.so
libhqdn3d.so
libils.so
libmiscfilters.so
libmvtools.so
libneo-dfttest.so
libneo-f3kdb.so
libneo-fft3d.so
libnnedi3.so
libremapframes.so
libremovegrain.so
libretinex.so
libsangnom.so
libtcanny.so
libtedgemask.so
libttempsmooth.so
libvsnlm_cuda.so
libvsznedi3.so

v0.3.0

We are excited to announce the v0.3.0 release of vs-pytorch~

• All-in-one CI build: remove CircleCI
• use torch 2.7.1+cu128 (50s' support)
• use FFmpeg 8.0

Breaking Changes

• remove havsfunc and Jet packages
• deprecate ROCm support

Base Environment

• system: Ubuntu 22.04
• GCC/G++: 13
• python: 3.10
• FFmpeg: 8.0
• VapourSynth: R70
• x264: latest
• x265: 4.1
• svt-av1-psy: latest
• aom: latest
• libvpx: latest
• fdk-aac: latest
• libass: latest

AI Environment

• CUDA toolkit: 12.9
• PyTorch: 2.7.1+cu128
• cupy: cuda12x

VapourSynth Python Plugin List

mbfunc
ccrestoration
ccvfi
vsutil
mvsfunc

VapourSynth C++ Plugin List

bestsource.so
libaddgrain.so
libakarin.so
libassrender.so
libbilateral.so
libbm3dcpu.so
libbm3dcuda.so
libbm3dcuda_rtc.so
libboxblur.so
libbwdif.so
libcas.so
libctmf.so
#libd2vsource.so
libdctfilter.so
libdescale.so
libdfttest2_cpu.so
libdfttest2_cuda.so
libdfttest2_nvrtc.so
libeedi2.so
libeedi3m.so
libffms2.so
libfillborders.so
libfluxsmooth.so
libfmtconv.so
libhqdn3d.so
libils.so
libmiscfilters.so
libmvtools.so
libneo-dfttest.so
libneo-f3kdb.so
libneo-fft3d.so
libnnedi3.so
libremapframes.so
libremovegrain.so
libretinex.so
libsangnom.so
libtcanny.so
libtedgemask.so
libttempsmooth.so
libvsnlm_cuda.so
libvsznedi3.so

v0.2.0

• Base Image: vs-ffmpeg-docker:v0.1.0

---

• mbfunc v0.1.0
• ccrestoration v0.2.1
• ccvfi v0.0.1

We are excited to announce the v0.2.0 release of vs-pytorch, which includes significant improvements and optimizations to enhance performance and reduce docker image size:

• Two-stage build: build ffmpeg and encoders at vs-ffmpeg-docker in GitHub Actions, then build vs-pytorch and vs-playground in CircleCI

• Removal of Conda: we have removed Conda from the Docker image

• Self-Compiled Encoders and Filters: all encoders and filters are now self-compiled, ensuring optimal compatibility and performance within the Docker environment

• Self-Compiled Common VapourSynth Filters: we have compiled a set of commonly used VapourSynth filters to provide a more comprehensive and ready-to-use environment for video processing

• AI Optimization: optimized AI inference speed, and support experimental video frame interpolation by ccvfi

• Our VapourSynth function library: mbfunc

Environment Details:

- system: Ubuntu 22.04
- GCC/G++: 13
- python: 3.10
- FFmpeg: latest
- VapourSynth: R70
- x264: latest
- x265: 4.1
- svt-av1-psy: latest
- aom: latest
- libvpx: latest
- fdk-aac: latest
- libass: latest

VapourSynth Python Plugin List

mbfunc
ccrestoration
ccvfi
vsutil
mvsfunc
havsfunc
JET's funcs

VapourSynth C++ Plugin List

bestsource.so
libaddgrain.so
libakarin.so
libassrender.so
libbilateral.so
libbm3dcpu.so
libbm3dcuda.so
libbm3dcuda_rtc.so
libboxblur.so
libbwdif.so
libcas.so
libctmf.so
libd2vsource.so
libdctfilter.so
libdescale.so
libdfttest2_cpu.so
libdfttest2_cuda.so
libdfttest2_nvrtc.so
libeedi2.so
libeedi3m.so
libffms2.so
libfillborders.so
libfluxsmooth.so
libfmtconv.so
libhqdn3d.so
libils.so
libmiscfilters.so
libmvtools.so
libneo-dfttest.so
libneo-f3kdb.so
libneo-fft3d.so
libnnedi3.so
libremapframes.so
libremovegrain.so
libretinex.so
libsangnom.so
libtcanny.so
libtedgemask.so
libttempsmooth.so
libvsnlm_cuda.so
libvsznedi3.so

v0.1.1

• ccrestoration v0.2.1
• use torch 2.1.2+cu121
• feat: add ffmpeg param templates

v0.1.0

• ccrestoration v0.2.0

v0.0.2

• ccrestoration v0.1.2

v0.0.1

• ccrestoration v0.1.0