app.py

from transformers import pipeline
import gradio as gr

p = pipeline("automatic-speech-recognition", model="aware-ai/wav2vec2-base-german", feature_extractor = "aware-ai/wav2vec2-xls-r-1b-5gram-german")

def transcribe(audio):
    return p(audio, chunk_length_s=10, stride_length_s=(4, 2))["text"]

gr.Interface(
    fn=transcribe, 
    inputs=[
        gr.inputs.Audio(source="microphone", type="filepath")
    ],
    outputs=[
        "textbox"
    ]).launch(server_name = "0.0.0.0")