stream_gradio_audio.py

import gradio as gr
import math
import numpy as np
import time
import io
import wave


def wave_header_chunk(frame_input=b"", channels=1, sample_width=2, sample_rate=24000):
    # This will create a wave header then append the frame input
    # It should be first on a streaming wav file
    # Other frames better should not have it (else you will hear some artifacts each chunk start)
    wav_buf = io.BytesIO()
    with wave.open(wav_buf, "wb") as vfout:
        vfout.setnchannels(channels)
        vfout.setsampwidth(sample_width)
        vfout.setframerate(sample_rate)
        vfout.writeframes(frame_input)

    wav_buf.seek(0)
    return wav_buf.read()


def yield_from_stream():
    # Necessary Start of Audio Stream
    yield wave_header_chunk()

    # Dummy Audio Generation with Sin Waves
    volume = 10000  # range [0.0, 1.0]
    fs = 48000  # sampling rate, Hz, must be integer
    duration = 0.25  # in seconds, may be float
    f = 440.0  # sine frequency, Hz, may be float

    # generate samples, note conversion to float32 array
    num_samples = int(fs * duration)

    while True:
        if f == 440.0:
            f = 880.0
        else:
            f = 440.0
        
        # Generated Audio To Add to Stream
        samples = np.array(
            [
                int(volume * math.sin(2 * math.pi * k * f / fs))
                for k in range(0, num_samples)
            ],
            dtype=np.int16,
        )
        
        # Yield to Stream
        yield samples.tobytes()
        
        # Dummy Delay To Simulate Generation Time
        time.sleep(0.5)


with gr.Blocks() as demo:
    out = gr.Audio(
        streaming=True,
        autoplay=True,
        interactive=False,
    )
    start = gr.Button("Start Stream")

    start.click(yield_from_stream, [], [out])

demo.launch(share=True)