app.py

import gradio as gr

t = "the big blue " * 5
i = 0

def transcribe(audio: str, state: str):
    global i
    next_chunk = t[i:i+1]
    i += 1
    state += " " + next_chunk
    return audio, state

with gr.Blocks() as demo:
    gr.Markdown("# New Audio Streaming In 🎤")
    inp = gr.Audio(sources=["microphone"], type="filepath")
    out = gr.Audio(type="filepath", autoplay=True)
    transcription = gr.State(value="")
    clear = gr.Button("Clear")

    inp.stream(transcribe, [inp, transcription], [out, transcription], time_limit=10, stream_every=1)

if __name__ == "__main__":
    demo.launch()