Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/KoljaB/RealtimeSTT/llms.txt

Use this file to discover all available pages before exploring further.

FunASR is Alibaba DAMO’s industrial-grade speech recognition toolkit. It supports 50+ languages, speaker diarization, emotion detection, streaming inference, and runs at up to 170× real-time speed. It is especially well-suited for Chinese speech recognition through models like SenseVoiceSmall and Paraformer-zh, but works for many other languages as well.

​
Install

FunASR is not bundled with any RealtimeSTT extra. Install it directly: 
pip install funasr

​
Engine Name

Pass "funasr" as the transcription_engine parameter: 
transcription_engine="funasr"

​
Basic Usage

from RealtimeSTT import AudioToTextRecorder

recorder = AudioToTextRecorder(
    transcription_engine="funasr",
    model="iic/SenseVoiceSmall",
    device="cuda",
)

​
Model Selection

Known model names such as SenseVoiceSmall, Fun-ASR-Nano, and Paraformer-zh are downloaded automatically through ModelScope when first used. Pass the model name or a full ModelScope repository path: 
recorder = AudioToTextRecorder(
    transcription_engine="funasr",
    model="iic/SenseVoiceSmall",
    device="cuda",
)

​
Configuration Options

The following RealtimeSTT parameters map directly to FunASR AutoModel arguments:
RealtimeSTT parameterFunASR mapping
modelmodel
devicedevice
beam_sizebeam_size
batch_sizebatch_size
transcription_engine_options: {"vad_filter": bool, "vad_model": str}vad_model

​
VAD Integration

To use FunASR’s built-in VAD model, pass vad_filter and vad_model together via transcription_engine_options: 
recorder = AudioToTextRecorder(
    transcription_engine="funasr",
    model="iic/SenseVoiceSmall",
    device="cuda",
    transcription_engine_options={
        "vad_filter": True,
        "vad_model": "fsmn-vad",
    },
)

​
Notes and Limitations

The FunASR integration is still under active development. If you encounter an issue, please open a GitHub issue on the RealtimeSTT repository.
For more details about FunASR itself, see the official FunASR GitHub repository.

Build docs developers (and LLMs) love

Get started for freeTalk to us