When using yield, automatic audio playback through web pages does not work properly

[AnitaSherry]

• I have searched to see if a similar issue already exists.
  yes

Is your feature request related to a problem? Please describe.
I hope the audio module can play multiple audio in sequence

Describe the solution you'd like
When I am using a large model for streaming output, I want to play the voice converted from text as well

Additional context
Generate audio code

async def generate_audio_from_text(text):
    tts = edge_tts.Communicate(text=text, voice='zh-CN-YunxiNeural', rate='-4%', volume='+0%', pitch='+0Hz')
    audio_bytes = b''
    sample_rate = 24000  # 设置采样率
    async for message in tts.stream():
        if message['type'] == 'audio':
            audio_bytes += message['data']
    
    # 将字节串转换为 NumPy 数组
    audio_data = np.frombuffer(audio_bytes, dtype=np.int16)

    return gr.Audio(value=(sample_rate, audio_data))

The code submitted to the webpage

    messages = history[-1][0]
    past_key_values = None
    current_length = 0
    current_sentence = ""
    
    for response, historys in model.stream_chat(tokenizer, messages, history=historys):
        new_token = response[current_length:]
        if new_token != '':
            history[-1][-1] += new_token
            current_sentence += new_token
            if contains_punctuation(new_token):
                yield history[-1][-1], await generate_audio_from_text(current_sentence)
                current_sentence = ""
            else:
                yield history[-1][-1], None
            current_length = len(response)

Streaming text can output normally, but audio cannot
Usually only the last audio will be output

[freddyaboulton]

Do you have autoplay=True set in the output audio? If you could share the full reproduction code, someone can take a closer look.

[AnitaSherry]

Do you have autoplay=True set in the output audio? If you could share the full reproduction code, someone can take a closer look.

yes，i set autoplay=True

with gr.Interface(
    predict,
    inputs=gr.Textbox(label="Input Text", lines=10),
    outputs=[
        gr.Textbox(label="Output Text"),
        gr.Audio(label="Output Audio", type="numpy", autoplay=True)
    ]
) as interface:
    
    interface.launch(server_name="xxx.xxx.xx.xx", server_port=xxxx, inbrowser=True, share=True)

I don't need to display the complete code to explain my problem, I just want to support continuous playback of multiple audio in a list

[abidlabs]

Closed via #8906. If you'd like to try it out, you can install gradio from this branch: #8843