import nemo.collections.asr as nemo_asr
import gradio as gr

asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name="QuartzNet15x5Base-En")


def transcribe(audio):
    transcription = asr_model.transcribe(paths2audio_files=[audio])
    return transcription[0]

interface = gr.Interface(fn=transcribe, inputs="audio", outputs="text")
interface.launch()