# FireRedTeam/FireRedASR

Open-source industrial-grade ASR models supporting Mandarin, Chinese dialects and English, achieving a new SOTA on public Mandarin ASR benchmarks, while also offering outstanding singing lyrics recognition capability.

Repository: https://github.com/FireRedTeam/FireRedASR
Canonical: https://ross.abutalabs.com/products/fireredasr
Language: Python
License: Apache-2.0
License Family: permissive
Topics: asr, industrial-grade, llm, multimodal-llm, open-source, speech-recognition, automatic-speech-recognition, conformer, speechllm, transformer
Last push: 2026-02-25T07:14:23+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 69, release rhythm 35, longevity 41
- inputs: {"age_days": 586, "days_push": 189, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1971, forks 165 (observed 2026-08-28T04:06:00.925898+00:00)

## What it is
FireRedASR is a family of open-source industrial-grade automatic speech recognition models supporting Mandarin, Chinese dialects, and English, with an LLM-based variant (Encoder-Adapter-LLM) and an efficient AED variant. It achieves state-of-the-art results on public Mandarin ASR benchmarks and offers strong singing lyrics recognition, with model weights and Python inference code released under Apache-2.0.

## Use cases
- transcribe Mandarin audio to text
- recognize speech in Chinese dialects
- transcribe English speech
- extract lyrics from songs
- build voice assistants with speech-to-text
- run high-accuracy ASR on GPU locally
- benchmark ASR models on aishell datasets

## When to choose
- you need SOTA Mandarin or Chinese dialect transcription accuracy
- you need lyrics recognition from singing audio
- you want open-source ASR weights you can self-host
- you can run GPU inference and want an LLM-based speech model

## When to avoid
- you need lightweight CPU-only or real-time on-device ASR
- you need languages beyond Mandarin, Chinese dialects, and English
- you need a full production ASR pipeline with VAD, LID, and punctuation (see FireRedASR2S)
- you need streaming low-latency transcription out of the box

## Facets
- artifact type: library
- maturity: active
- function: speech-recognition, machine-learning, llm-inference, nlp
- domain: speech-processing, machine-learning, artificial-intelligence
- platform: python
- tags: asr, mandarin, chinese-dialects, speechllm, conformer, lyrics-recognition, model-weights, inference, natural-language-processing, gpu, linux

## Member repositories
- FireRedTeam/FireRedASR (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:00.925898+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 2026-08-30T03:05:03.417327+00:00, confidence not recorded.
  - readme: https://github.com/FireRedTeam/FireRedASR (fetched 2026-08-28T04:06:00.925898+00:00, sha 30b40356ecf2)
- Data as of 2026-08-30T08:39:29.467469+00:00.