Fetching contributors…
Cannot retrieve contributors at this time
116 lines (82 sloc) 3.44 KB


The unit and integration tests are contained in the aeneas.tests subpackage.

Unit and Integration

The tests cover both the library aeneas and the CLI tools

Currently, there are more than 1,200 tests.

Structure of the Tests

Each test file roughly corresponds to the source file being tested, with the exception of a couple of modules, whose tests have been split across multiple test files.

The following naming convention is used:

  • regular tests are specified in files with prefix test_;
  • tests requiring network access have prefix net_test_;
  • tests with long running time have prefix long_test_;
  • tests included in the benchmark suite have prefix bench_test_;
  • tests for have prefix tool_test_.

Running the Tests

The test harness is the standard Python library unittest, hence no additional testing libraries/frameworks should be needed in order to run the tests.

Tests are not distributed via PyPI, since some of them require large audio or data files. If you installed aeneas using PyPI (e.g., via pip), you cannot run the tests. You must git clone the main repository instead.

Certain tests might fail if you do not set your environment up properly, for example, if you do not have optional libraries/tools installed (e.g., festival or speect, etc.).

In the root directory of the repository there is a script that helps running the unit/integration tests. Run:

$ # get the usage
$ python -h

$ # run all regular tests (< 3 minutes)
$ python

$ # run all tool tests (< 5 minutes)
$ python -t

$ # run all long tests (< 15 minutes)
$ python -l

$ # run all benchmark tests (< 15 minutes)
$ python -b

$ # run all regular tests with verbose output including error messages
$ python -v

$ # run all network tests (requires an Internet connection)
$ python -n

$ # run only textfile tests
$ python textfile

$ # run only run_vad tool tests
$ python -t run_vad

The venvs directory contains some scripts to automate the creation and management of virtual enviroments for testing purposes.

Release Process

Before releasing a new version of aeneas, all tests (regular, tools, long, benchmark, network) are run on all supported platforms and Python interpreters.

Any failure blocks the release process, until the failure is removed.

Tests might be run on other (i.e., not officially supported) platforms or interpreters as well, but any failure on them does not automatically block the release process.

Please consider sponsoring the project to add more extensive tests or to include your platform in the release testing process. See the README file for details.


Performance tests are located in a separate repository, with the corresponding documentation.

The benchmark results can be consulted online at the following URL: