Skip to content

ReleaseNotes.md

Latest commit

 

History

History
414 lines (317 loc) · 24.8 KB

ReleaseNotes.md

File metadata and controls

414 lines (317 loc) · 24.8 KB
Raw

NVEnc Release Notes

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

9.23

  • Add libvmaf CUDA support. ( #781 )
  • 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 --avs input.
  • 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 degrain fallback analyze MV/SAD snapshot for scene-change delayed output.
  • Fix candidate list in --vpp-denoise-dct.

9.22

9.21

9.20

9.19

  • Bundle nnedi3_weight.bin in the package.
  • Fix vpp-kfm RTGMC preset medium error (vpp-nnedi nnsize=5) by reducing NNEDI register usage. ( #776 )
  • Fix --vpp-rtgmc chroma_motion=false processing. ( #777 )
  • Fix failed to copy edi side-data frame with chroma_motion=true, source_match=3.
  • Fix Degrain analysis result reuse in --vpp-rtgmc.
  • Fix frame pool allocation in --vpp-rtgmc-search-prefilter.

9.18

  • Add new high quality deinterlace filter --vpp-kfm which supports 24/30/60 mixed VFR. ( #677 )
  • Add new high quality deinterlace filter --vpp-rtgmc. ( #161, #218, #267, #677)
  • Add --vpp-deint-csp to specify CSP for deinterlace filters.
  • Add new filter --vpp-degrain.　( #161, #218, #267, #677)
  • Update --vpp-nnedi to new specification.
  • Support odd crop values when output resolution must be even. ( #772 )
  • Fix CUDA texture border handling in --vpp-yadif.
  • Fix audio desync when libavformat returns negative pts.

9.17

9.16

  • Improve subtitle burn-in for Blu-ray and MPEG-TS inputs. ( #756 )
  • Fix libopus encoding for 5.1 / 7.1 channel layouts.

9.15

9.14

  • Add --vmaf support for Linux systems. ( #755 )
  • Improve vapoursynth error messages.

9.13

9.12

  • Fix for DTS:X not being copyable.

9.11

  • Add option to append input command line parameters to encoding_tool in muxer metadata. (--muxer-add-cmd)
  • Fix potential SIGPIPE(141) error termination in Linux multi-GPU environment.
  • Fix error on finalization when encoding E-AC3. ( #706 )
  • Avoid mixing GPUs with different B-frame availability when using --parallel.
  • Migrate Linux build to meson.

9.10

  • Add feature to set --audio-bitrate to different value depending on audio channels. ( #743 )

9.09

  • Add support for Vapoursynth API V4. ( #747 )

9.08

9.07

  • Fix error when encoding H.264 for RTMP/FLV output.
  • Fix mkv output failure when encoding with -c av_libsvtav1. ( #733 )
  • Add option to show preset/tune parameters (--check-preset-params).

9.06

  • Fix --vpp-resize bilinear,spline*,lanczos* creating artifacts depending on resize ratio. ( #698, #737 )

9.05

9.04

  • Improve DX11 device initialization to not detect virtual/remote adaptors. ( #725 )
  • Improve progress indicator when using --parallel.
  • Add support for using --parallel with multiple pipes.

9.03

  • Fix 9.02 did not incude NVEncNVSDKNGX.dll. ( #724 )

9.02

  • Fix error when using both ngx-vsr and ngx-true-hdr together (issue since 9.00). ( #724 )
  • Avoid unintended fps values when front of input file is corrupted.

9.01

  • Improve handling when input files have negative pts.
  • Improve quality of burned in subtitles in --vpp-subburn processing by changing libass initialization method. ( #717 )

9.00

  • Add NVEnc.auo2 with native support for AviUtl2.
  • Add feature to use filters with avcodec encoders.
    • Available with -c av_xxx Example: -c av_libsvtav1 --avcodec-prms "preset=6,crf=30,svtav1-params=enable-variance-boost=1:variance-boost-strength=2" Other usable options include av_libvvenc, av_libvpx-vp9, etc.
  • Update ffmpeg libraries. (Windows)
    • ffmpeg 7.1+ (20240822) -> 8.0
    • libpng 1.6.44 -> 1.6.50
    • expat 2.6.2 -> 2.7.1
    • fribidi 1.0.11 -> 1.0.16
    • libogg 1.3.5 -> 1.3.6
    • libxml2 2.12.6 -> 2.14.5
    • libvpl 2.13.0 -> 2.15.0
    • libvpx 1.14.1 -> 1.15.2
    • dav1d 1.4.3 -> 1.5.1
    • libxxhash 0.8.2 -> 0.8.3
    • glslang 15.0.0 -> 15.4.0
    • dovi_tool 2.1.2 -> 2.3.1
    • libjpeg-turbo 2.1.0 -> 3.1.1
    • lcms2 2.16 -> 2.17
    • zimg 3.0.5 -> 3.0.6
    • libplacebo 7.349.0 -> 7.351.0
    • libsvtav1 3.1.0 (new!) x64 only
    • libvvenc 1.13.1 (new!) x64 only
    • libass 0.9.0 -> 0.17.4 (x64), 0.14.0 (x86)
    • harfbuzz 11.4.4 (new)
    • libunibreak 6.1 (new)
    • Remove mmt/tlv patch

8.11

  • Added options to change CUDA optimization mode (--cuda-stream, --cuda-mt, #710)
    • To address CUDA_ERROR_MAP_FAILED which seems to occur on RTX50xx, multi-threaded calls to the same CUDA context are disabled by default (--cuda-mt 0)
  • Fixed --vpp-rff not working properly in NVEnc 8.10.
  • Add option for --bitstream-padding for AV1 CBR encoding ( #714 ).

8.10

  • Fix filtering crushing when using with interlaced encoding.

8.09

  • Updates for NVEnc.auo (AviUtl/AviUtl2 plugin).

8.08

8.07

  • Fix issues with raw output when using formats like yuv4mpegpipe. ( #699 )
  • Fix potential freeze when using raw output.
  • Add support for --option-file on Linux.

8.06

  • Fix performance degradation caused in 8.05. ( #696 )

8.05

  • Add support for combining --output-format with -c raw. ( #693 ) Now supports cases like -c raw --output-format nut.
  • Fix black/white processing in 10-bit depth for --vpp-edgelevel.
  • Improve interlace detection when using --avsw. ( #688 )

8.04

8.03

  • Improve audio and video synchronization to achieve more uniform mixing when muxing with subtitles or data tracks.
  • Improve invalid input data hadling to avoid freeze when "failed to run h264_mp4toannexb bitstream filter" error occurs. Now properly exits with error.
  • Add support for uyvy as input color format. ( #678 )
  • Fix application freezing when using readers other than avhw.
  • Automatically disable --parallel when number of encoders is 1 when using --parallel auto.

8.02

  • Fix vpp-resize ngx-vsr, libplaceo* not working in 8.01. ( #683 )

8.01

  • Fix insufficient frame buffer causing error termination when using readers other than avhw (issue since 8.00beta1).
  • Fix crash on process termination in Linux environment (issue since 8.00beta1).
  • Fix hw decode not working in Linux environment (issue since 8.00beta1).
  • Improve stability of Vulkan initialization in Linux environment.
  • Avoid unnecessary Dolby Vision RPU conversion.
  • Add detailed logging for errors during Dolby Vision RPU conversion.
  • Update documentation.

8.00beta7

8.00beta6

  • Add parallel encoding feature with file splitting. (--parallel)
  • Add support for ISO 639-2 T-codes in language code specification. ( #674 )
  • Continue processing even when DirectX11/Vulkan initialization fails. ( #675 )
  • Fix timestamps occasionally becoming incorrect when using --seek with certain input files.
  • Increase priority of GPUs with multiple encoders in auto GPU selection.
  • Fix potential freeze when encoder output thread encounters an error.
  • Fix potential freeze when encoder terminates with an error.
  • Fix incorrect handling of pts for frames before keyframe when decoding from middle of OpenGOP encoded files.

8.00beta5

8.00beta4

  • Fix some codecs not being able to decode with avsw since 8.00b2.
  • Fix interlaced encoding not working when using filters since 8.00b1.
  • Add 10.0, 10.1, 10.2, 10.4 options to --dolby-vision-profile. ( #672 )

8.00beta3

  • Fix --dolby-vision-profile not working with readers other than avhw/avsw. ( #663 )
  • Fix memory allocation failure when using yuv422 output. ( #670 )

8.00beta2

  • Improve auto gpu selection for multi (NVIDIA) GPU environments.

8.00beta1

  • Add support for NVENC SDK 13.0. Driver 570.00 or later is required.
    • add support for H.264/AV1 in --tf-level.
    • Add forced_4 to --split-enc.
    • Support HEVC --temporal-layers.
    • Add experimental support for yuv422 output. (but untested, as I have no Blackwell GPUs)
  • Update CUDA for Windows build to CUDA 11.8.
    • Requires driver 452.39 or later.
    • Requires CC3.5 or later GPUs.
      • NVIDIA GPUs GeForce Maxwell generation or later and some Kepler (GTX Titan, 780(Ti), 730-710)
    • GPUs that do not support CC3.5 (GTX770, 760, 740, 6xx) are no longer supported.
  • No longer supports builds with CUDA 10 or earlier.
    • Builds for Ubuntu 18.04 removed.
  • Noe NVEnc will be able to handle "frame transfer from CPU to GPU", "filtering", and "frame submission to encoder" in parallel.
  • Merge AV1 --max-temporal-layers to --temporal-layers to be same as other codecs.
  • Improve dolby vision rpu handling. ( #663 )

7.82

  • Now AV1 --level can be specified again.
  • Added checks for --max-bitrate and --ref to avoid errors when specifying --level.
  • Avoid --qvbr not working properly when setting --max-bitrate too big. ( #486 )
  • Fix some case that audio not being able to play when writing to mkv using --audio-copy.
  • Now more detailed encoder error information will be shown.

7.81

  • Avoid width field in mp4 Track Header Box getting 0 when SAR is undefined. ( #680 )

7.80

7.79

  • Fix documents for --vpp-resize nvvfx-superres ( #658 ).
  • Fix --trim being offset for a few frames when input file is a "cut" file (which does not start from key frame) and is coded using OpenGOP.