# pykaldi/pykaldi

A Python wrapper for Kaldi

Repository: https://github.com/pykaldi/pykaldi
Canonical: https://ross.abutalabs.com/products/pykaldi
Homepage: https://pykaldi.github.io
Language: Python
License: Apache-2.0
License Family: permissive
Topics: python, wrapper, kaldi, openfst, asr, speech-recognition, speech, language-model, feature-extraction, clif, numpy
Last push: 2025-11-30T20:55:37+00:00

## Health v2 (maintenance only)
Score: 47/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 54, release rhythm 8, longevity 100
- inputs: {"age_days": 3362, "days_push": 276, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1039, forks 247 (observed 2026-08-28T04:03:19.847862+00:00)

## What it is
PyKaldi is a Python scripting layer providing wrappers for the C++ APIs of the Kaldi speech recognition toolkit and OpenFst library. It enables low-level access to Kaldi functions, objects, and data structures from Python, plus high-level modules for ASR, alignment, and segmentation.

## Use cases
- run a pre-trained Kaldi ASR model from Python
- extract speech features like MFCCs and ivectors in Python
- manipulate Kaldi lattices and OpenFst FSTs in Python code
- convert Kaldi matrices to NumPy arrays and back
- prototype new speech recognition algorithms without writing C++
- implement custom Kaldi tools in Python
- force-align and segment audio transcripts

## When to choose
- you need low-level access to Kaldi or OpenFst internals from Python
- you are a speech recognition researcher building custom pipelines
- you want to integrate a trained Kaldi system into a Python application
- you need to read, write, or manipulate Kaldi file formats and tables

## When to avoid
- you just want a simple pretrained speech-to-text API with no Kaldi knowledge
- you need a maintained, actively developed library with broad community support
- you want end-to-end deep learning ASR like Whisper rather than classic Kaldi/HMM-GMM systems
- you cannot build C++ dependencies, since installation is complex

## Facets
- artifact type: library
- maturity: maintenance
- function: speech-recognition, nlp, audio-processing, machine-learning, parser
- domain: speech-processing, machine-learning, developer-tools
- platform: python, cpp
- tags: kaldi, openfst, asr, wrapper, clif, numpy, feature-extraction, language-model, natural-language-processing, linux, macos

## Member repositories
- pykaldi/pykaldi (main) score 47

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:19.847862+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:03:55.210496+00:00, confidence not recorded.
  - readme: https://github.com/pykaldi/pykaldi (fetched 2026-08-28T04:03:19.847862+00:00, sha a0bd84914a79)
  - homepage: https://pykaldi.github.io (fetched 2026-08-29T13:04:59.699226+00:00, sha f583adf6bb29)
- Data as of 2026-08-30T08:39:29.467469+00:00.