app.py

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

def transcribe(state, file):

  if not audio:
    return {state_var: state, transcription_var: state} 
  asr_model = nemo_asr.models.EncDecCTCModelBPE.from_pretrained(
              model_name="stt_rw_conformer_ctc_large")
                
  transcription= asr_model.transcribe([audio])
  return  transcription[0]


with gr.Blocks() as demo:
  state_var = gr.State("")

  with gr.Row():
    with gr.Column():
      uploaded_audio = gr.Audio(label="Upload Audio File", type="filepath", source="upload")

    with gr.Row():
      transcribe_button = gr.Button("Transcribe")

    transcribe_button.click(
      transcribe,
      [uploaded_audio],
    )

  
demo.launch()