The EAT is a set of tools for processing ADM (Audio Definition Model) files. It can convert ADM files between profiles, validate them, render them, fix common issues, and more.
It contains:
- A framework for building processing graphs that operate on audio and associated data.
- A set of ADM-related processes that sit within this framework.
- A command-line tool
eat-process, which processes ADM files using the processes from the framework, as defined in a configuration file. - A set of example configuration files for the tool, for doing things like profile conversion, loudness measurement, fixing common issues, validation etc.
For more information, see the documentation at https://ebu-adm-toolbox.readthedocs.io/
The eat-process tool requires only one argument: the configuration file which
specifies the processes to run and how to connect them.
eat-process /path/to/configuration_file.jsonConfiguration files can specify everything required to perform some processing, but normally they leave parameters such as paths to input and output files unspecified. These can be passed on the command-line with the -o option, which takes a process and parameter name, and a value to set it to. For example:
eat-process example_configs/conform_block_timing.json -o input.path in.wav -o output.path out.wavThis will enforce a minimum block duration on in.wav, writing to out.wav.
For information on the config file format, options and available processes, see the documentation at https://ebu-adm-toolbox.readthedocs.io/
To compile the toolbox you will require a working C++20 development environment. It's currently tested on ubuntu-latest, macos-latest and windows-latest; see here for details.
CMake version 3.21 or higher is required to configure and build the project.
If building via vcpkg/the supplied CMake presets, the Ninja build system is required along with the following:
// on macOS/linux
pkg-config
// on linux (macOS has these by default)
curl
zip
unzip
tar
To build:
git clone [email protected]:ebu/ebu-adm-toolbox.git --recursive
cd ebu-adm-toolbox
# This step may not be neccessary and should only be required once
./external/vcpkg/bootstrap-vcpkg.sh # bootstrap_vcpkg.bat if on windows
# This step configures and builds using vcpkg, see CMakePresets.json for details
cmake --preset release && cmake --build --preset releaseReplace the git@... url with the https equivalent if not using ssh, and replace release with debug if a debug build is required.
On Windows a 64 bit developer command prompt should be used, or the preset can be built directly from Visual Studio using its inbuilt CMake support.
The CMake preset uses vcpkg to build the project's dependencies, alternatively these can be manually built by the user
libadm v0.13
libbw64 v0.10
libear v0.9
libebur128 v1.2.6
cli11 v2.2.0
nlohmann/json v3.11.2
valijson v1.0
Catch2 v3.0.1 // Only required for tests
Try and adhere to the suggestions in the pitchfork layout, using 'separate header placement' with merged tests (i.e. separate src/include paths but tests in src).
The library is called 'eat' for EBU ADM Toolbox, it comprises:
- framework (graph generation and execution)
- processes (i/o, various metadata and audio manipulation)
- render (BS.2127 renderer implementation)
- utilities (support code hared between the above These components are split into eat::framework, eat::process etc. namespaces, also reflected in the directory structure
- Use CPM.cmake for header only libraries, cmake config or small / quick to build dependencies?
- Document dependencies and attempt to find in exported config but do not force any management system on users?
- Use clang format and cmake formatting via Format.cmake
- Catch2 v3 built as a static library and using minimal includes (start with catch/test_macros.hpp)
- Use built in cmake for test registration
- Start with one test binary for binary, probably add separate ones for algorithms as we add them
- export a cmake config for ease of inclusion
Apache 2.0, see LICENSE.txt.