Runnable examples

These scripts mirror the main README snippets and are intended to run from a fresh clone. Install the base requirements first:

pip install -r requirements.txt

Run commands from the repository root. Each script defaults to the bundled example/zh.mp3 audio inside the downloaded model and accepts a custom audio path when you want to test your own file.

Quickstart with funasr.AutoModel

python examples/quickstart.py --hub hf --device cuda:0
python examples/quickstart.py path/to/audio.wav --language 中文 --hotwords 开放时间

Direct model inference

Use this when you want to call model.py directly without the high-level AutoModel wrapper.

python examples/direct_inference.py --device cuda:0
python examples/direct_inference.py path/to/audio.wav

Speaker diarization

This example enables VAD, CAM++ speaker labels, and punctuation restoration.

python examples/speaker_diarization.py --hub hf --device cuda:0
python examples/speaker_diarization.py path/to/meeting.wav

vLLM offline batch inference

Install vLLM before running this example:

pip install "funasr>=1.3.3" "vllm>=0.12.0"
python examples/vllm_batch.py --tensor-parallel-size 1
python examples/vllm_batch.py audio1.wav audio2.wav --hotwords 张三 北京

AutoModelVLLM decodes each input in a single pass. For long meetings, segment the audio first or use the AutoModel(..., vad_model="fsmn-vad") path.

Streaming SDK

Install vLLM before running this example:

pip install "funasr>=1.3.3" "vllm>=0.12.0"
python examples/streaming_sdk.py --hub hf --chunk-ms 720
python examples/streaming_sdk.py path/to/audio.wav