Releases: rigaya/NVEnc
Releases · rigaya/NVEnc
Release list
9.32
- Update ffmpeg libraries. -> binaries and src, build_scripts
- ffmpeg 8.0 -> 9.0.1
- libvmaf 3.0.0 -> 3.2.0
- libpng 1.6.50 -> 1.6.58
- xz 5.8.2 -> 5.8.3
- expat 2.7.1 -> 2.8.2
- harfbuzz 11.4.4 -> 14.3.0
- libunibreak 6.1 -> 7.0
- libass 0.17.4 -> 0.17.5
- lame 3.100 -> 4.0
- libxml2 2.14.5 -> 2.15.3
- libbluray 1.3.4 -> 1.5.0
- libaribcaption 1.1.1 -> 1.1.2
- libav1d 1.5.3 -> 1.5.4
- libvpl 2.16.0 -> 2.17.0
- nv-codec-headers 12.2.72.0 -> 13.1.15.0
- glslang 15.4.0 -> 16.5.0
- shaderc 2024.1 -> 2026.3
- dovi_tool 2.3.1 -> 2.3.3
- libjpeg-turbo 3.1.1 -> 3.2.0
- lcms2 2.17 -> 2.19.1
- vulkan-loader 1.3.295 -> 1.4.359
- libplacebo 7.351.0 -> 7.360.1
- vvenc 1.13.1 -> 1.14.0
- svt-av1 3.1.0 -> 4.2.0
- dovi_tool 2.3.1 -> 2.3.3
- Reinitialize host path after --vpp-onnx zero-copy failure. ( #791 )
- Fix --vpp-anime4k-shader deblur strength depending on parameter order.
- Fix LUT3D input domain handling in --vpp-colorspace.
- Fix double application of saturation in --vpp-tweak.
9.31
- Update to NVENC SDK 13.1.
- Support AV1 hierarchical B-frame reference mode in --bref-mode.
- Support mid-stream resolution changes.
- Add --adapt-resolution to set maximum mid-stream input resolution changes.
- Add support to track PMT change.
- Add ONNX deinterlace filter --vpp-onnx-deint.
- Support language exclusion for audio/subtitle track selection (--audio-copy / --sub-copy).
- Improve --vpp-onnx zero-copy path to reduce memory usage. ( #791 )
- Fix GPU memory growth in long --vpp-kfm mode=24 runs.
- Improve memory retention for --vpp-kfm and NVENC input.
- Output mp4 trailer even if error has occurred.
9.30
9.29
9.28
- Optimize --vpp-kfm / --vpp-degrain.
- Fix libvmaf CUDA implementation and access violation. ( #788 )
- Update Linux package CUDA to 11.8.
- Optimize --vpp-onnx. (Thanks @teaching-droid!)
- Better performance by adding with zero-copy path.
- Support frames setting of registered models.
- Better model handling.
Contributors
teaching-droid
Assets 8
9.27
- Optimize --vpp-kfm / --vpp-rtgmc / --vpp-degrain by adding SAD-threshold early termination for motion search.
- Optimize --vpp-kfm by removing redundant processing.
- Handle mixed RFF sources properly in --vpp-kfm.
- Fix --vpp-kfm mode=24 stall and timestamps.
- Fix chroma degrain analysis in --vpp-rtgmc. ( #786 )
- Fix high bit-depth variance calculation in --vpp-nnedi. ( #779 )
9.26
- Add --vpp-lenscorrection and --vpp-v360.
- Add mask input and multi-frame (temporal) support to --vpp-onnx.
- Add
customparameter support to --vpp-libplacebo-shader. ( #764 ) - Fix --vpp-onnx fp16 being treated as fp32 with TensorRT.
- Optimize --vpp-kfm / --vpp-degrain.
- Fix and optimize --vpp-rtgmc.
- Fix 10-bit overflow in --vpp-rtgmc. ( #779 )
- Add statistics to --vship-ssimulacra2 log output.
- Separate TensorRT caches by runtime environment and input conditions.
9.25
- Add Windows named pipe support. ( #785 )
- Improve accuracy of --vpp-kfm.
- Speed up --vpp-degrain / --vpp-rtgmc with zero-copy cache references and parallelized motion search.
- Update libvmaf and enable float. ( #783 )
- Add automatic CPU fallback for libvmaf. ( #783 )
- Fix --vpp-libplacebo-tonemapping rejecting bool strings for
use_dovi. - Fix potential issues.
9.24
- Improve --vpp-kfm performance for long processing.
- Extend temporal options in --vpp-fft3d.
- Improve diagnostics when ONNX Runtime fails to load.
- Avoid nan in CUDA --vmaf. ( #781 )
- Fix Linux VMAF CUDA support. ( #781 )
- Enable libvship build on Linux.
- Add --vpp-rife-ov RIFE v4.x frame interpolation filter.
- Add SDR to HDR model support to --vpp-onnx.
- New model files can be downloaded from the link ( https://github.com/rigaya/HWEnc-onnx-models/releases ).
- Clarify error for non-existent VMAF json model. ( #783 )
- Add OpenVINO RIFE frame interpolation filter --vpp-rife-ov.
- Add coring and hue range to --vpp-tweak.
- Add interpolation mode and fix all in --vpp-curves.
- Add detailed parameters to --vpp-hqdering.
- Add chroma to --vpp-cas.
- Add valid source size to --vpp-descale.
- Add keep and fix state management in --vpp-mpdecimate.
- Add detection parameters to --vpp-ivtc.
- Add planes to --vpp-nnedi.
- Add temporal radius and improve temporal processing in --vpp-knn.
- Fix yuv444→y410 conversion.
- Fix mask output in --vpp-msmooth.
- Fix chroma processing in --vpp-vinverse.
- Propagate Yadif copyFrameAsync error in --vpp-yadif.
- Fix nvvfx output crop height check.
- Fix alpha input buffer width.
- Fix smooth QP table reference coordinate clamp.
- Fix Anime4K spline36 linear term in --vpp-anime4k-shader.
- Fix AFS mie_spot8 4th element reference in --vpp-afs.
- Fix colorspace LUT3D green axis scale in --vpp-colorspace.
- Fix libplacebo radius warning argument in --vpp-libplacebo-shader.
- Fix libplacebo deband grain target in --vpp-libplacebo-deband.
- Fix delogo two-point least squares intercept in --vpp-delogo.
- Fix decimate diff grid height in --vpp-decimate.
- Fix decomb stripe grid Y stride in --vpp-decomb.
- Fix chroma reference line in RGB→YUV420 conversion.
- Apply denoise filter fixes.
- Fix threshold in --vpp-convolution3d.
9.23
- Add libvmaf CUDA support for
--vmaf. ( #781 )
Allows much faster VMAF calculation. - Improve precision of --vpp-finedehalo. ( #777 )
- Mark --vpp-finedehalo as interlace unsupported. ( #782 )
- Fix frame corruption in --vpp-ivtc expand/mixed path.
- Fix chroma joint-bilateral OOB in --vpp-anime4k-shader on YUV444.
- Fix SRCPTR clamp upper limit in --vpp-afs map_filter.
- Fix OOB write in --vpp-subburn when subtitle position is negative.
- Fix --vpp-deflicker disabled after scene change.
- Fix audio track selection fallback loop in
--avsinput. - Fix LUT boundary clamp and spline coefficient reference in --vpp-curves.
- Fix constant-luminance YUV→RGB conversion using wrong inverse matrix in --vpp-colorspace.
- Fix YUV444/NV24 color conversion bugs.
- Fix --vpp-degrain causing ghost in some inputs. ( #778 )
- Fix candidate list in --vpp-denoise-dct.
