FunASR

Overview

FunASR is an open-source speech recognition toolkit in Python. It supports training, inference, streaming ASR, VAD, punctuation, speaker diarization, and serving via OpenAI-compatible or MCP interfaces.

Best for

Best for
Developers building production speech recognition systems who need an integrated open-source toolkit

Use cases

• Building speech-to-text pipelines with VAD and punctuation
• Fine-tuning ASR models on custom audio datasets
• Deploying streaming ASR services through OpenAI-compatible endpoints

Notes

FunASR is an open-source speech recognition toolkit in Python. It supports training, inference, streaming ASR, VAD, punctuation, speaker diarization, and serving via OpenAI-compatible or MCP interfaces.

20,579 stars on GitHub. Last updated 2026-10-02. Licensed MIT.

Use cases

• Building speech-to-text pipelines with VAD and punctuation
• Fine-tuning ASR models on custom audio datasets
• Deploying streaming ASR services through OpenAI-compatible endpoints

Pros

• Comprehensive all-in-one toolkit for speech recognition tasks
• Large community with over 20,000 GitHub stars
• Supports modern serving protocols like OpenAI-compatible and MCP

Cons

• Steep learning curve due to many components and configuration options
• May require significant compute resources for training and diarization
• Documentation can be scattered across multiple sources

Indexed from awesome-llmops and enriched against its public facts.