Home

Awesome

media-autobuild_suite

Before opening an issue, check if it's an issue directly from executing the suite. This isn't Doom9, reddit, stackoverflow or any other forum for general questions about the things being compiled. This script builds them, that's all.

This source code is also mirrored in GitLab.

Most git sources in the suite use GitHub, so if it's down, it's probably useless to run the suite at that time.

Download

Click here to download latest version

For information about the compiler environment see the wiki, there you also have a example of how to compile your own tools.

Included Tools And Libraries

Information about FFmpeg external libraries


Requirements



Information


This tool is inspired by the very nice, linux cross-compiling tool from Roger Pack (rdp): https://github.com/rdp/ffmpeg-windows-build-helpers

It is based on msys2 and tested under Windows 8.1, 10 and 11. http://sourceforge.net/projects/msys2/

I use some jscript parts from nu774: https://github.com/nu774/fdkaac_autobuild

Thanks to all of them!

This Windows Batchscript setups a Mingw-w64/GCC compiler environment for building ffmpeg and other media tools under Windows. After building the environment it retrieves and compiles all tools. All tools get static compiled, no external .dlls needed (with some optional exceptions)

How to use it:

The script writes an .ini file at /build/media-autobuild_suite.ini, so you only need to make these choices the first time what you want to build.

The script doesn't build any registry key or system variables, when you don't need it any more you can delete the folder and your system will be clean. Building everything from scratch takes about ~3 hours depending on how many CPU cores are utilized and what is enabled.

Check forcing-recompilations for documentation on how you can force a rebuild of all libs/binaries.

To save a bit of space after compiling, you can delete all source folders (except the folders with a "-git" or "-svn" on end) in /build. There's an option in the .bat for the script to remove these folders automatically. To save even more space, you can delete /msys64 after compiling. If the suite is run after /msys64 has been deleted, it will download again.

Have fun!

Troubleshooting


If there's some error during compilation follow these steps:

  1. Make sure you're using the latest version of this suite by downloading the latest version and replacing all files with the new ones;
  2. If you know which part it's crashing on, delete that project's folder in /build and run the script again (ex: if x264 is failing, try deleting x264-git folder in /build);
  3. If it still doesn't work, create an issue and paste the URL to logs.zip that the script gives or attach the file yourself to the issue page.
  4. If the problem isn't reproducible by the contributors of the suite, it's probably a problem on your side. Delete /msys64 and /local[32|64] if they exist. /build is usually safe to keep and saves time;
  5. If the problem is reproducible, it could be a problem with the package itself or the contributors will find a way to probably make it work.
  6. If you compile with --enable-libnpp and/or --enable-cuda-nvcc, see Notes about CUDA SDK

What The Individual Files Do


media-autobuild_suite.bat

/build/media-autobuild_suite.ini

/build/media-suite_compile.sh

/build/media-suite_update.sh

/build/media-suite_helper.sh

/build/media-suite_deps.sh

/build/ffmpeg_options.txt & /build/mpv_options.txt

Optional User Files


/local32|64/etc/custom_profile & $HOME/custom_build_options

Custom Patches


Using custom patches is not officially supported, if you do use custom patches, do not expect much support from this suite. Either go to the patch's author or to a forum for help


To reference the cloned folder itself (flac-git ffmpeg-git) you can use the variable ${REPO_DIR}. To reference the generated build folder, you would need to use ${REPO_DIR}/build-${bits} to reference the build-64bit folder for 64 bit builds or build-32bit folder for 32 bit builds.

If you are building for both 32 and 64-bit, you can gate certain commands using if [[ $bits = 64bit ]]; then <command>; fi to only run them when building for 64 bits.
Be careful with certain packages due to their unique build process (x265, x264, and some others) so make sure to check the compile script

Example Script: /build/aom_extra.sh for aom-git

#!/bin/bash

# Don't automatically run cmake || configure
touch do_not_reconfigure

# Commands to run before running cmake
_pre_cmake(){
    # Installs libwebp
    do_pacman_install libwebp
    # Downloads the patch and then applies the patch
    do_patch "https://gist.githubusercontent.com/1480c1/9fa9292afedadcea2b3a3e067e96dca2/raw/50a3ed39543d3cf21160f9ad38df45d9843d8dc5/0001-Example-patch-for-learning-purpose.patch"
    # Change directory to the build folder
    cd_safe "build-${bits}"
    # Run cmake with custom options. This will override the previous cmake commands.
    # $LOCALDESTDIR refers to local64 or local32
    cmake .. -G"Ninja" -DCMAKE_INSTALL_PREFIX="$LOCALDESTDIR" \
        -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang \
        -DBUILD_SHARED_LIBS=off -DENABLE_TOOLS=off
}

_post_cmake(){
# post cmake and post configure will be unavailable due to "touch do_not_reconfigure"
# as the do_not_reconfigure flag will skip the post commands.
}

# Commands to run before building using ninja
_pre_ninja(){
    # Change directory to the build folder (Absolute path or relative to aom-git)
    cd_safe "build-${bits}"
    # applies a local patch (Absolute or relative to aom-git)
    do_patch "My-Custom-Patches/test-diff-files.diff"
    # run a custom ninja command.
    ninja aom_version_check
    # Not necessary, but just for readability sake
    cd_safe ..
}

Example Script: /build/ffmpeg_extra.sh for ffmpeg-git

#!/bin/bash

# Force to the suite to think the package has updates to recompile.
# Alternatively, you can use "touch recompile" for a similar effect.
touch custom_updated

_pre_configure(){
    #
    # Apply a patch from ffmpeg's patchwork site.
    do_patch "https://patchwork.ffmpeg.org/patch/12563/mbox/" am
    #
    # Apply a local patch inside the directory where is "ffmpeg_extra.sh"
    patch -p1 -i "$LOCALBUILDDIR/ffmpeg-0001-my_patch.patch"
    #
    # Add extra configure options to ffmpeg (ffmpeg specific)
    # If you want to add something to ffmpeg not within the suite already
    # you will need to install it yourself, either through pacman
    # or compiling from source.
    FFMPEG_OPTS+=(--enable-libsvthevc)
    #
}

_post_make(){
    # Don't run configure again.
    touch "$(get_first_subdir -f)/do_not_reconfigure"
    # Don't clean the build folder on each successive run.
    # This is for if you want to keep the current build folder as is and just recompile only.
    touch "$(get_first_subdir -f)/do_not_clean"
}

For a list of possible directive, look under unset_extra_script in media-suite_helper.sh. Beware as they may change in the future.

Notes about CUDA SDK


This is for cuda-nvcc and libnpp, not for NVENC, it is built with ffmpeg by default

For --enable-cuda-nvcc and --enable-libnpp to work, you need NVIDIA's CUDA SDK installed with CUDA_PATH variable to be set system-wide (Usually set by default on CUDA SDK install) and VS2017 or better installed which should come with vswhere.exe.
If for some reason CUDA_PATH isn't set and/or vswhere.exe isn't installed, you need to export the CUDA_PATH variable path using the above mentioned user files and manually export the correct PATH including the absolute cygpath converted path to MSVC's cl.exe.

You do not need to do the following if you installed the SDK with the default locations etc and you have 8.3 short paths enabled or if you installed to a directory without any spaces

If you did not understand any of the words above, assume the best and hope the compilation will succeed the first time, else try reinstalling the SDK and MSVC to a path without any spaces.

You will only need to be worried if running the following command in the mintty terminal produces a path with a space or if you have 8.3 short paths disabled.

cygpath -sm "$CUDA_PATH"

If running the above command produces a path with a space, you will need to either disable cuda/npp stuff or reinstall your cuda sdk to a path without spaces.

Nothing should be disabled manually when installing CUDA SDK as disabling random things can cause the compilation to fail

For example, if you need to manually set the CUDA_PATH and include in the PATH the binaries for MSVC cl.exe and nvcc.exe, add this bit of bash script inside a text file in /local64/etc/custom_profile:

# adapt these to your environment
_cuda_basepath="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA"
_cuda_version=10.0

_msvc_basepath="C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC"
_msvc_version=14.15.26726
_msvc_hostarch=x64
_msvc_targetarch=x64

# you shouldn't need to change these unless your environment is weird or you know what you're doing
export CUDA_PATH=$(cygpath -sm "${_cuda_basepath}")/${_cuda_version}
export PATH=$PATH:$(dirname "$(cygpath -u "\\${_msvc_basepath}\\${_msvc_version}\bin\Host\\${_msvc_hostarch}\\${_msvc_targetarch}\cl.exe")")
export PATH=$PATH:$CUDA_PATH/bin