app.py

import streamlit as st
import edge_tts
import asyncio
import librosa
import soundfile
import io

from inference.infer_tool import Svc

def get_or_create_eventloop():
    try:
        return asyncio.get_event_loop()
    except RuntimeError as ex:
        if "There is no current event loop in thread" in str(ex):
            loop = asyncio.new_event_loop()
            asyncio.set_event_loop(loop)
            return asyncio.get_event_loop()

def tts_get_voices_list():
    voices = []
    tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
    for item in tts_voice_list:
        voices.append(item['ShortName'])
    
    return voices

loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)

with st.form(key = 'tts', clear_on_submit=False):
    txt = st.text_input('your text message (your text message should be < 100 characters)', max_chars = 100)
    voice = str(st.selectbox('voices', tts_get_voices_list()))
    summitted = st.form_submit_button('Summit')

    if summitted:
        tts = asyncio.run(edge_tts.Communicate(txt, voice).save('temp\\test.mp3'))
        audio, sr = librosa.load('temp\\test.mp3', sr=16000, mono=True)
        raw_path = io.BytesIO()
        soundfile.write(raw_path, audio, 16000, format="wav")
        raw_path.seek(0)
        model = Svc(fr"Herta-Svc/G_10000.pth", f"Herta-Svc/config.json", device = 'cpu')
        out_audio, out_sr = model.infer('speaker0', 0, raw_path, auto_predict_f0 = True,)
        soundfile.write('temp\\xxx.wav', out_audio.cpu().numpy(), 44100)
        audio_file = open('temp\\xxx.wav', 'rb')
        audio_bytes = audio_file.read()
        st.audio(audio_bytes, format = 'audio/wav')