app.py

import IPython

import sys
import subprocess

subprocess.check_call([sys.executable, "-m", "pip", "install", "--upgrade", "--force-reinstall", "git+https://github.com/osanseviero/tortoise-tts.git"])

# entmax could not be installed at same time as torch
subprocess.check_call([sys.executable, "-m", "pip", "install", "entmax"])

from tortoise_tts.api import TextToSpeech
from tortoise_tts.utils.audio import load_audio, get_voices
import torch 
import torchaudio
import gradio as gr

device = "cuda" if torch.cuda.is_available() else "cpu"

# This will download all the models used by Tortoise from the HF hub
tts = TextToSpeech(device="cuda")

voices = [
  "angie",
  "daniel",
  "deniro",
  "emma",
  "freeman",
  "geralt",
  "halle",
  "jlaw",
  "lj",
  "snakes",
  "William",
]
voice_paths = get_voices()
print(voice_paths)

preset = "fast"

def inference(text, voice):
    text = text[:256]
    cond_paths = voice_paths[voice]
    conds = []
    print(voice_paths, voice, cond_paths)
    for cond_path in cond_paths:
        c = load_audio(cond_path, 22050)
        conds.append(c)
    print(text, conds, preset)
    gen = tts.tts_with_preset(text, conds, preset)
    print("gen")
    torchaudio.save('generated.wav', gen.squeeze(0).cpu(), 24000)
    return "generated.wav"
    

 
text = "Joining two modalities results in a surprising increase in generalization! What would happen if we combined them all?"
examples = [
    [text, "angie"],
    [text, "emma"],
    ["how are you doing this day", "freeman"]
]

block = gr.Blocks()
with block:
    gr.Markdown("# TorToiSe")
    gr.Markdown("A multi-voice TTS system trained with an emphasis on quality")
    with gr.Tabs():
      with gr.TabItem("Pre-recorded voices"):
        iface = gr.Interface(
          inference,
          inputs=[
              gr.inputs.Textbox(type="str", default=text, label="Text", lines=3),
              gr.inputs.Dropdown(voices),
          ],
          outputs="audio",
          enable_queue=True,
          examples=examples,
        )

    gr.Markdown("This demo shows the ultra fast option in the TorToiSe system. For more info check the <a href='https://github.com/neonbjb/tortoise-tts' target='_blank'>Repository</a>.",)

block.launch()


iface.launch(cache_examples=True)