# facebookresearch/omnilingual-asr

Omnilingual ASR Open-Source Multilingual SpeechRecognition for 1600+ Languages

Repository: https://github.com/facebookresearch/omnilingual-asr
Canonical: https://ross.abutalabs.com/products/omnilingual-asr
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-12-30T20:48:26+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 59, release rhythm 62, longevity 21
- inputs: {"age_days": 300, "days_push": 246, "days_rel": 254, "gap_med": 9, "n_releases_24m": 2}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2898, forks 269 (observed 2026-08-28T04:07:28.796922+00:00)

## What it is
An open-source multilingual speech recognition library from Meta AI supporting over 1,600 languages, including hundreds never previously covered by ASR. It provides a family of CTC and LLM-based models (300M to 7B parameters) with zero-shot and few-shot adaptation for adding new languages from just a few paired examples.

## Use cases
- transcribe speech in low-resource languages
- add speech recognition for a new language with few examples
- batch transcribe long audio recordings
- build multilingual voice applications
- research zero-shot speech recognition
- convert speech to text for 1600+ languages

## When to choose
- you need ASR coverage for rare or low-resource languages
- you want state-of-the-art multilingual transcription accuracy
- you need to adapt speech recognition to a new language without large datasets
- you need unlimited-length audio decoding

## When to avoid
- you only need a handful of high-resource languages and a lighter-weight model
- you lack GPU resources for the larger 3B/7B checkpoints
- you need fine-tuning recipes for the unlimited-length model variant
- you need a permissively licensed model for commercial use without reviewing the license

## Facets
- artifact type: library
- maturity: active
- function: speech-recognition, audio-processing, machine-learning, deep-learning, transformers
- domain: speech-processing, machine-learning
- platform: python, cross-platform
- tags: asr, speech-to-text, multilingual, zero-shot-learning, wav2vec2, ctc, llm-asr, transcription, low-resource-languages, meta-ai, natural-language-processing, audio, gpu

## Member repositories
- facebookresearch/omnilingual-asr (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:28.796922+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 2026-08-30T07:35:12.776311+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/omnilingual-asr (fetched 2026-08-28T04:07:28.796922+00:00, sha 08cd6ff8978e)
  - registry_pypi: https://pypi.org/pypi/omnilingual-asr/json (fetched 2026-08-29T09:49:56.481855+00:00, sha 3312058b7cbf)
- Data as of 2026-08-30T08:39:29.467469+00:00.