app.py

import gradio as gr

from audio_streaming_client import AudioStreamingClient

audio_streaming_client = AudioStreamingClient()
audio_streaming_client.start()


def stream_audio(audio):
    sample_rate = audio[0]
    audio_streaming_client.put_audio(audio[1], sample_rate)
    output_size = len(audio[1])
    output_audio = audio_streaming_client.get_audio(sample_rate, output_size)
    return (sample_rate, output_audio)


with gr.Blocks() as demo:
    gr.Markdown("# Speech to speech in an inference endpoint 🎤")
    inp = gr.Audio(sources=["microphone"], type="numpy")
    out =  gr.Audio(streaming=True, autoplay=True)
    inp.stream(stream_audio, inp, out, time_limit=600, stream_every=1)

demo.launch()