app.py

import torch
import os
import streamlit as st
from TTS.api import TTS
from tempfile import NamedTemporaryFile

# By using XTTS you agree to CPML license https://coqui.ai/cpml
os.environ["COQUI_TOS_AGREED"] = "1"

def generate_audio(audio_file, text_input):
  # Initialize model
  model = "tts_models/multilingual/multi-dataset/xtts_v2"
  device = 'cuda' if torch.cuda.is_available() else 'cpu'
  tts = TTS(model).to(device)

  with NamedTemporaryFile(delete=False, suffix='.wav') as tmp_file:
    output_path = tmp_file.name
    tts.tts_to_file(text=text_input, speaker_wav=audio_file, language='en', file_path=output_path)

  return output_path

def main():
  # Title
  title = f"""<h1 align="center" style="font-size: 2rem";>Voice Clone</h1>"""
  st.markdown(title, unsafe_allow_html=True)

  # Subtitle
  title = f"""<h2 align="center" style="font-size: 1.2rem; margin-bottom: 2rem;">Make your favorite characters say anything!</h2>"""
  st.markdown(title, unsafe_allow_html=True)

  # Upload audio file
  uploaded_file = st.file_uploader('Add an audio file of the voice you want to clone...', type=['wav'])

  if uploaded_file is not None:
    reference_audio, synthesized_audio = st.columns(2)
    with reference_audio:
      st.header('Reference Audio')
      st.audio(uploaded_file, format='audio/wav')

    # Input text 
    text_input = st.text_input('What do you want your character to say? (Alphabet letters only, DO NOT INCLUDE PUNCTUATION)')

    if st.button('Synthesize'):
      if text_input:
        try:
          with st.spinner('Synthesizing...'):
            output_path = generate_audio(uploaded_file, text_input)

          with synthesized_audio:
            st.header('Synthesized Audio')
            st.audio(output_path, format='audio/wav')
        except:
          st.error('There was an issue synthesizing the text. Please check the input and try again. Remember, do not include punctuation.')
      else:
        st.error('Please provide a text input!')

if __name__ == '__main__':
  main()