- Linux docker image (
akashisn/ffmpeg
) & binary release7.0.2
6.1.2
5.1.6
4.4.5
- Windows binary release
7.0.2
6.1.2
5.1.6
4.4.5
-
ffmpeg
-
linux/amd64
configure options:
--enable-zlib --enable-lzma --enable-gmp --enable-iconv --enable-gnutls --enable-libsrt --enable-libopenjpeg --enable-libwebp --enable-libvpx --enable-libx264 --enable-libx265 --enable-libaom --enable-libvmaf --enable-libopus --enable-libvorbis --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libvo-amrwbenc --enable-libmp3lame --enable-libfreetype --enable-libfribidi --enable-libxml2 --enable-libfontconfig --enable-libass --enable-libaribb24 --enable-sdl2 --enable-cuda-llvm --enable-ffnvcodec --enable-cuvid --enable-nvdec --enable-nvenc --enable-libdrm --enable-vaapi --enable-libvpl --disable-autodetect --disable-debug --disable-doc --enable-gpl --enable-version3 --extra-libs='-lm -lpthread -lstdc++' --pkg-config-flags=--static --prefix=/usr/local
Dependent library
$ ldd ffmpeg linux-vdso.so.1 (0x00007ffed66ba000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f38fcc95000) libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f38fca69000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f38fc841000) /lib64/ld-linux-x86-64.so.2 (0x00007f3901805000) libmvec.so.1 => /lib/x86_64-linux-gnu/libmvec.so.1 (0x00007f38fc744000) libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f38fc724000)
-
-
Windows ffmpeg
-
windows/x64
configure options:
--enable-zlib --enable-libopenjpeg --enable-libwebp --enable-lzma --enable-gmp --enable-iconv --enable-gnutls --enable-libsrt --enable-libvpx --enable-libx264 --enable-libx265 --enable-libaom --enable-libopus --enable-libvorbis --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libvo-amrwbenc --enable-libmp3lame --enable-libfreetype --enable-libfribidi --enable-libxml2 --enable-libfontconfig --enable-libass --enable-libaribb24 --enable-sdl2 --enable-cuda-llvm --enable-ffnvcodec --enable-cuvid --enable-nvdec --enable-nvenc --enable-libmfx --enable-d3d11va --enable-dxva2 --arch=x86_64 --cross-prefix=x86_64-w64-mingw32- --disable-autodetect --disable-debug --disable-doc --disable-w32threads --enable-cross-compile --enable-gpl --enable-version3 --extra-libs='-static -static-libgcc -static-libstdc++ -Wl,-Bstatic -lm -lpthread -lstdc++' --extra-cflags=--static --target-os=mingw64 --pkg-config=pkg-config --pkg-config-flags=--static -prefix=/usr/local
Dependent library
$ LANG=C objdump -p ffmpeg.exe | grep 'DLL Name:' DLL Name: ADVAPI32.dll DLL Name: bcrypt.dll DLL Name: GDI32.dll DLL Name: KERNEL32.dll DLL Name: msvcrt.dll DLL Name: ole32.dll DLL Name: OLEAUT32.dll DLL Name: PSAPI.DLL DLL Name: SHELL32.dll DLL Name: SHLWAPI.dll DLL Name: USER32.dll DLL Name: AVICAP32.dll DLL Name: WS2_32.dll
-
https://trac.ffmpeg.org/wiki/HWAccelIntro
$ ffmpeg -hide_banner -hwaccels
Hardware acceleration methods:
cuda
vaapi
qsv
drm
-
cuda
: NVIDIA's GPU accelerated video codecs (nvenc/nvdec
)$ ffmpeg -hide_banner -encoders | grep nvenc V....D av1_nvenc NVIDIA NVENC av1 encoder (codec av1) V....D h264_nvenc NVIDIA NVENC H.264 encoder (codec h264) V....D hevc_nvenc NVIDIA NVENC hevc encoder (codec hevc)
-
vaapi
: Intel Media Driver for VAAPI$ ffmpeg -hide_banner -encoders | grep vaapi V....D h264_vaapi H.264/AVC (VAAPI) (codec h264) V....D hevc_vaapi H.265/HEVC (VAAPI) (codec hevc) V....D mjpeg_vaapi MJPEG (VAAPI) (codec mjpeg) V....D mpeg2_vaapi MPEG-2 (VAAPI) (codec mpeg2video) V....D vp8_vaapi VP8 (VAAPI) (codec vp8) V....D vp9_vaapi VP9 (VAAPI) (codec vp9)
-
QSV
: Intel QSV (Intel Quick Sync Video)$ ffmpeg -hide_banner -encoders | grep qsv V..... av1_qsv AV1 (Intel Quick Sync Video acceleration) (codec av1) V..... h264_qsv H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (Intel Quick Sync Video acceleration) (codec h264) V..... hevc_qsv HEVC (Intel Quick Sync Video acceleration) (codec hevc) V..... mjpeg_qsv MJPEG (Intel Quick Sync Video acceleration) (codec mjpeg) V..... mpeg2_qsv MPEG-2 video (Intel Quick Sync Video acceleration) (codec mpeg2video) V..... vp9_qsv VP9 video (Intel Quick Sync Video acceleration) (codec vp9)
https://trac.ffmpeg.org/wiki/Hardware/QuickSync
You can use the following command to find out which codecs are supported by your CPU.
$ sudo docker run --rm -it --device=/dev/dri akashisn/vainfo
Trying display: drm
libva info: VA-API version 1.19.0
libva info: User environment variable requested driver 'iHD'
libva info: Trying to open /usr/local/lib/dri/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_19
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.19 (libva 2.19.0)
vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 23.3.3 (1c13afa)
vainfo: Supported profile and entrypoints
VAProfileNone : VAEntrypointVideoProc
VAProfileNone : VAEntrypointStats
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Simple : VAEntrypointEncSlice
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointFEI
VAProfileH264Main : VAEntrypointEncSliceLP
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointFEI
VAProfileH264High : VAEntrypointEncSliceLP
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileJPEGBaseline : VAEntrypointVLD
VAProfileJPEGBaseline : VAEntrypointEncPicture
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264ConstrainedBaseline: VAEntrypointFEI
VAProfileH264ConstrainedBaseline: VAEntrypointEncSliceLP
VAProfileVP8Version0_3 : VAEntrypointVLD
VAProfileVP8Version0_3 : VAEntrypointEncSlice
VAProfileHEVCMain : VAEntrypointVLD
VAProfileHEVCMain : VAEntrypointEncSlice
VAProfileHEVCMain : VAEntrypointFEI
VAProfileHEVCMain10 : VAEntrypointVLD
VAProfileHEVCMain10 : VAEntrypointEncSlice
VAProfileVP9Profile0 : VAEntrypointVLD
VAProfileVP9Profile2 : VAEntrypointVLD
VAEntrypointEncSlice
: Can encodeVAEntrypointVLD
: Can decode
You can find the pre-built binary files on the release page.
There are files in the assets with the following naming conventions:
ffmpeg-${version}-${"linux" or "windows"}-${arch}.tar.gz
Contains of archive file:
$ ls
bin configure_options lib run.sh
$ ls bin/
ffmpeg ffplay ffprobe
$ tree lib/
lib
├── dri
│ ├── i965_drv_video.so
│ └── iHD_drv_video.so
├── libdrm.so -> libdrm.so.2
├── libdrm.so.2 -> libdrm.so.2.123.0
├── libdrm.so.2.123.0
├── libdrm_amdgpu.so -> libdrm_amdgpu.so.1
├── libdrm_amdgpu.so.1 -> libdrm_amdgpu.so.1.123.0
├── libdrm_amdgpu.so.1.123.0
├── libdrm_intel.so -> libdrm_intel.so.1
├── libdrm_intel.so.1 -> libdrm_intel.so.1.123.0
├── libdrm_intel.so.1.123.0
├── libdrm_nouveau.so -> libdrm_nouveau.so.2
├── libdrm_nouveau.so.2 -> libdrm_nouveau.so.2.123.0
├── libdrm_nouveau.so.2.123.0
├── libdrm_radeon.so -> libdrm_radeon.so.1
├── libdrm_radeon.so.1 -> libdrm_radeon.so.1.123.0
├── libdrm_radeon.so.1.123.0
├── libigdgmm.so -> libigdgmm.so.12
├── libigdgmm.so.12 -> libigdgmm.so.12.5.0
├── libigdgmm.so.12.5.0
├── libigfxcmrt.so -> libigfxcmrt.so.7
├── libigfxcmrt.so.7 -> libigfxcmrt.so.7.2.0
├── libigfxcmrt.so.7.2.0
├── libmfx-gen
│ └── enctools.so
├── libmfx-gen.so -> libmfx-gen.so.1.2
├── libmfx-gen.so.1.2 -> libmfx-gen.so.1.2.12
├── libmfx-gen.so.1.2.12
├── libmfx.so -> libmfx.so.1
├── libmfx.so.1 -> libmfx.so.1.35
├── libmfx.so.1.35
├── libmfxhw64.so -> libmfxhw64.so.1
├── libmfxhw64.so.1 -> libmfxhw64.so.1.35
├── libmfxhw64.so.1.35
├── libpciaccess.so -> libpciaccess.so.0
├── libpciaccess.so.0 -> libpciaccess.so.0.11.1
├── libpciaccess.so.0.11.1
├── libva-drm.so -> libva-drm.so.2.2200.0
├── libva-drm.so.2 -> libva-drm.so.2.2200.0
├── libva-drm.so.2.2200.0
├── libva.so -> libva.so.2.2200.0
├── libva.so.2 -> libva.so.2.2200.0
├── libva.so.2.2200.0
└── mfx
├── libmfx_h264la_hw64.so
├── libmfx_hevc_fei_hw64.so
├── libmfx_hevcd_hw64.so
├── libmfx_hevce_hw64.so
├── libmfx_vp8d_hw64.so
├── libmfx_vp9d_hw64.so
└── libmfx_vp9e_hw64.so
3 directories, 49 files
$ cat run.sh
#!/bin/sh
export PATH=$(dirname $0)/bin:$PATH
export LD_LIBRARY_PATH=$(dirname $0)/lib:$LD_LIBRARY_PATH
export LIBVA_DRIVERS_PATH=$(dirname $0)/lib/dri
export LIBVA_DRIVER_NAME=iHD
exec $@
If you use run.sh
, you can run it after setting the LD_LIBRARY_PATH
and other settings.
And, if you want to encode with QSV, you need to run it with root privileges.
sample:
$ sudo ./run.sh ffmpeg -y \
-init_hw_device qsv:hw \
-hwaccel qsv \
-hwaccel_output_format qsv \
-extra_hw_frames 32 \
-i https://github.com/bower-media-samples/big-buck-bunny-1080p-60fps-30s/raw/master/video.mp4 \
-c:v h264_qsv \
-f mp4 \
video-h264_qsv.mp4
When running in Docker, you need to mount the DRI device.
sample:
$ sudo docker run --rm -it --device=/dev/dri -v `pwd`:/workdir \
akashisn/ffmpeg:7.0.2 -y \
-loglevel verbose \
-init_hw_device qsv:hw \
-hwaccel qsv \
-hwaccel_output_format qsv \
-extra_hw_frames 32 \
-i video.mp4 \
-c:v h264_qsv \
-f mp4 \
video-h264_qsv.mp4
If you want to use a non-free codec(e.g. fdk-aac
, libnpp
), you can generate a binary in the current directory by executing the following command.
Generated binaries cannot be redistributed due to licensing issues. Please use them for your own use only.
# for windows build
> $Env:CUDA_SDK_VERSION = "12.0.1"
> $Env:NVIDIA_DRIVER_VERSION = "528.33"
> curl -L -o cuda_${Env:CUDA_SDK_VERSION}_${Env:NVIDIA_DRIVER_VERSION}_windows.exe https://developer.download.nvidia.com/compute/cuda/${Env:CUDA_SDK_VERSION}/local_installers/cuda_${Env:CUDA_SDK_VERSION}_${Env:NVIDIA_DRIVER_VERSION}_windows.exe
> docker buildx build --build-arg HOST_TARGET=x86_64-w64-mingw32 --build-arg TARGET_OS=windows --build-arg CUDA_SDK_VERSION=${Env:CUDA_SDK_VERSION} --build-arg NVIDIA_DRIVER_VERSION=${Env:NVIDIA_DRIVER_VERSION} --output type=local,dest=build -t ffmpeg-nonfree:windows -f ./nonfree.Dockerfile .
# for linux build
$ touch cuda_11.6.0_511.23_windows.exe # dummy file
$ docker buildx build --output type=local,dest=build -t ffmpeg-nonfree:linux -f ./nonfree.Dockerfile .
QSV (Quick Sync Video) was previously supported by libmfx.so
provided by MediaSDK, but since the development of MediaSDK has been discontinued, it has been taken over by the successor, oneAPI's oneVPL (libvpl.so
).
In FFmpeg, the option --enable-libvpl
was added from version 6.0. Therefore, in this build, only version 6.0
supports QSV through oneVPL's dispatcher, and previous versions are built using the old MediaSDK dispatcher.
Note that the MediaSDK may no longer be usable as it lacks support for new hardware.
Moreover, oneVPL supports older hardware by calling MediaSDK
.
Therefore, unless there is a reason to use an older version of FFmpeg, it is recommended to use the version compatible with oneVPL.
graph TD;
VPL[oneVPL Dispatcher]-->oneVPL-intel-gpu;
VPL[oneVPL Dispatcher]-->MediaSDK;
https://github.com/oneapi-src/oneVPL/tree/master#onevpl-architecture
To execute QSV (Quick Sync Video) on a Virtual Machine, it is necessary to pass through Intel's integrated GPU (iGPU) to the VM.
Pass-through technologies include Intel GVT-g, SR-IOV, etc., and the compatibility varies depending on the generation of the CPU1.
For Intel GVT-g, please refer to the ArchWiki2.
In Proxmox, if you are using systemd-boot instead of GRUB, kernel parameters can be set using /etc/kernel/cmdline
. Also, don't forget to apply the changes by running proxmox-boot-tool refresh
. If necessary, adding kvm.ignore_msrs=1
is recommended3.
Using https://github.com/yugr/Implib.so, the dynamic library related to QSV is delay-loaded, so when you run ldd
to check, it appears to have no dependencies. However, please note that it becomes necessary when using QSV. This resolves the inconvenience where FFmpeg wouldn't even start without the dynamic library even when QSV is not needed, achieving a pseudo-static link. Therefore, we have consolidated the tags which were previously separated based on whether QSV was available or not.