app.py

from transformers import pipeline
import gradio as gr

pipe = pipeline(model="bhuang/wav2vec2-xls-r-1b-cv9-fr")

def transcribe(audio, state=""):
    text = pipe(audio, chunk_length_s=5, stride_length_s=1)["text"]
    state += text + " "
    return state, state

# streaming mode
iface = gr.Interface(
    fn=transcribe, 
    inputs=[
        gr.Audio(source="microphone", type="filepath", streaming=True, label="Record something..."),
        "state"
    ], 
    outputs=[
        "textbox",
        "state"
    ],
    title="Realtime ASR in French",
    # description="Realtime demo for French ASR using a fine-tuned wav2vec2 model.",
    allow_flagging="never",
    live=True
)

iface.launch()