Kaldi

Kaldi is a toolkit designed for speech recognition researchers and professionals. Written in C++ and licensed under Apache License v2.0, Kaldi aims to provide a modern and flexible code that is easy to modify and extend. Key features include code-level integration with Finite State Transducers (FSTs), extensive support for linear algebra using standard BLAS and LAPACK routines, and an extensible design that allows algorithms to work with various sources of scores, like neural networks. Kaldi is open-source, allowing modifications and re-releases. It offers complete recipes for building speech recognition systems using widely available databases.