Kaldi

Kaldi is an open source toolkit for speech recognition. Its code is published on GitHub, and the project site points visitors to the repository, the documentation, a models page and a contact for help. The toolkit is written in C++ and is built from source rather than installed as a packaged application. The repository includes build instructions for Unix systems including Linux, Darwin and Cygwin, plus separate instructions for Windows. Example system builds are provided as recipes in an examples directory, which shows researchers and engineers how to train recognizers on standard speech datasets. The documentation covers the techniques used, a tutorial on C++ coding and a reference for the code. A models page lists the pretrained model builds that are available, which the site itself describes as not many. Kaldi is for speech researchers and developers who want to build and train automatic speech recognition systems and are comfortable compiling and modifying C++ code. Support is through two mailing lists, one for users and one for developers, and contributions follow a standard GitHub fork and pull request process. The toolkit is free to download from its public repository, and the site lists no paid plans.