py-kaldi-asr and pykaldi
These are competitors offering different approaches to Python-Kaldi integration: py-kaldi-asr prioritizes simplicity and online decoding convenience with active maintenance (reflected in its monthly downloads), while pykaldi provides comprehensive but lower-level bindings to Kaldi's full API.
About py-kaldi-asr
gooofy/py-kaldi-asr
Some simple wrappers around kaldi-asr intended to make using kaldi's (online) decoders as convenient as possible.
Provides Python bindings for Kaldi's nnet3-chain and GMM online decoders via Cython, enabling real-time speech recognition directly within Python applications on Linux. Supports streaming audio decoding and likelihood scoring, integrating with pre-trained models from the Zamia Speech project. Requires manual pkg-config setup to link against Kaldi's C++ libraries and dependencies like OpenFST and ATLAS.
About pykaldi
pykaldi/pykaldi
A Python wrapper for Kaldi
Provides low-level bindings to Kaldi and OpenFst C++ libraries, enabling direct manipulation of acoustic models, lattices, and FST structures from Python. Includes specialized modules for feature extraction, neural network decoding, language modeling, and alignment—designed to complement Kaldi's command-line tools rather than replace them. Integrates NumPy for matrix operations and supports reading/writing Kaldi archives and table formats.
Scores updated daily from GitHub, PyPI, and npm data. How scores work