app.py

import streamlit as st
import edge_tts
import asyncio
import tempfile
import os
from typing import Dict
from collections import defaultdict

async def text_to_speech(text: str, voice: str) -> str:
    output_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
    communicate = edge_tts.Communicate(text, voice)
    await communicate.save(output_file.name)
    return output_file.name

async def list_voices() -> Dict[str, Dict]:
    voices = await edge_tts.list_voices()
    return {v['ShortName']: {'name': v['ShortName'], 'language': v['Locale']} for v in voices}

def process_voices(voices: Dict[str, Dict]) -> Dict[str, Dict[str, str]]:
    processed_voices = defaultdict(dict)
    for full_name, details in voices.items():
        language = details['language']
        speaker_name = full_name.split('-')[2].replace('Neural', '')
        processed_voices[language][speaker_name] = full_name
    return dict(processed_voices)

async def main():
    st.title("OpenSpeech TTS")
    st.write("An OpenAI compatible API to reproduce high fidelity speech fast, in minimal hardware")
    st.write("Official Repo: https://github.com/PantelisDeveloping/openspeech-tts/tree/main")

    # Get voices and process them
    voices = await list_voices()
    processed_voices = process_voices(voices)

    # Text-to-Speech
    st.header("Text-to-Speech")
    text_input = st.text_area("Enter text to convert to speech:")

    # Two-step voice selection
    col1, col2 = st.columns(2)
    with col1:
        selected_language = st.selectbox("Select language:", list(processed_voices.keys()))
    with col2:
        selected_speaker = st.selectbox("Select speaker:", list(processed_voices[selected_language].keys()))

    selected_voice = processed_voices[selected_language][selected_speaker]

    if st.button("Generate Speech"):
        if not text_input:
            st.error("Please enter some text.")
        else:
            with st.spinner("Generating speech..."):
                output_file = await text_to_speech(text_input, selected_voice)
                st.audio(output_file, format='audio/mp3')
                os.unlink(output_file)  # Delete the temporary file

    # List Available Voices
    # st.header("Available Voices")
    # for language, speakers in processed_voices.items():
    #    st.subheader(language)
    #    st.write(", ".join(speakers.keys()))

if __name__ == '__main__':
    asyncio.run(main())