app.py

from transformers import pipeline
import gradio as gr
import base64


pipe = pipeline("automatic-speech-recognition", 
                model="openai/whisper-base.en",
                chunk_length_s=30, device="cpu")

def get_transcription(audio, encoded_audio):
    print(audio, encoded_audio)
    if encoded_audio != "":
        decoded_file_data = base64.b64decode(encoded_audio)
        with open("temp_audio.wav", "wb") as f:
            f.write(decoded_file_data)
        
        audio = "temp_audio.wav"
    else:
        audio = audio[1]
    print(audio)
    response = pipe(audio)["text"]
    return response

audio = gr.Audio(source="microphone", type="filepath", label="Audio")
transciption = gr.Textbox(label="Transcription")

iface = gr.Interface(fn=get_transcription, 
                     inputs=["audio", "text"], 
                     outputs=transciption.style(show_copy_button=True),
                     title="Speech Transcription")
    
iface.launch()