app.py

import os
os.system("pip install nemo_toolkit['all']")
import gradio as gr
import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.ASRModel.from_pretrained("nvidia/stt_zh_citrinet_1024_gamma_0_25")

def trans(audio_path):
    transcriptions = asr_model.transcribe([audio_path])
    print(transcriptions)
    return transcriptions

outputs = [gr.outputs.Textbox(label="stt result")] 

interface = gr.Interface(fn=trans, 
                         inputs=gr.Audio(source="upload", label="上传音频（长度小于60秒）", type="filepath"),
                         outputs=outputs,
                         enable_queue=True)
interface.launch(debug=True)

# gr.Interface.load("models/nvidia/stt_zh_citrinet_1024_gamma_0_25").launch()