SenseVoiceSmall / demo.py
游雁
update
72963f9
raw
history blame
No virus
612 Bytes
from funasr import AutoModel
from funasr.utils.postprocess_utils import rich_transcription_postprocess
model_dir = "FunAudioLLM/SenseVoiceSmall"
model = AutoModel(
model=model_dir,
vad_model="fsmn-vad",
vad_kwargs={"max_single_segment_time": 30000},
device="cuda:0",
hub="hf",
)
# en
res = model.generate(
input=f"{model.model_path}/example/en.mp3",
cache={},
language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech"
use_itn=True,
batch_size_s=60,
merge_vad=True, #
merge_length_s=15,
)
text = rich_transcription_postprocess(res[0]["text"])
print(text)