Text_to_Speech / app.py
Rozinamax's picture
Update app.py
20d4ba4
raw
history blame
2.77 kB
import os
import torch
import gradio as gr
import torchaudio
import time
from datetime import datetime
from tortoise.api import TextToSpeech
from tortoise.utils.audio import load_voice, load_voices
VOICE_OPTIONS = [
"angie",
"applejack",
"atkins",
"barack_obama",
"daniel",
"daws",
"deniro",
"dortice",
"dreams",
"emma",
"empire",
"freeman",
"geralt",
"grace",
"halle",
"jane_eyre",
"jlaw",
"kennard",
"lescault",
"lj",
"mol",
"mouse",
"myself",
"pat",
"pat2",
"rainbow",
"sanjita",
"snakes",
"tim_reynolds",
"tom",
"weaver",
"william",
"random",
]
def inference(
text,
voice,
Emotion,
Preset,
):
texts = [text]
Angry_tone = "[I am so angry]"
Sad_tone = "[I am so sad]"
Happy_tone = "[I am so happy]"
Scared_tone = "[I am so scared]"
if Emotion == "Angry":
text = Angry_tone + text
if Emotion == "Sad":
text = Sad_tone + text
if Emotion == "Happy":
text = Happy_tone + text
if Emotion == "Scared":
text = Scared_tone + text
voices = [voice]
if len(voices) == 1:
voice_samples, conditioning_latents = load_voice(voice)
else:
voice_samples, conditioning_latents = load_voices(voices)
audio_frames = []
for j, text in enumerate(texts):
for audio_frame in tts.tts_with_preset(
text,
voice_samples=voice_samples,
conditioning_latents=conditioning_latents,
preset=Preset,
k=1
):
audio_frames.append(torch.from_numpy(audio_frame.cpu().detach().numpy()))
complete_audio = torch.cat(audio_frames, dim=0)
yield (24000, complete_audio.numpy())
def main():
title = "TTS "
text = gr.Textbox(
lines=4,
label="Text:",
)
voice = gr.Dropdown(
VOICE_OPTIONS, value="jane_eyre", label="Select voice:", type="value"
)
Emotion = gr.Radio(
["Angry", "Sad", "Happy", "Scared"],
type="value",
)
Preset = gr.Radio(
["ultra_fast", "fast", "standard", "high_quality"],
type="value",
value="ultra_fast",
)
output_audio = gr.Audio(label="streaming audio:", streaming=True, autoplay=True)
interface = gr.Interface(
fn=inference,
inputs=[
text,
voice,
Emotion,
Preset,
],
title=title,
outputs=[output_audio],
)
interface.queue().launch()
if __name__ == "__main__":
tts = TextToSpeech(kv_cache=True, use_deepspeed=True, half=True)
with open("Tortoise_TTS_Runs_Scripts.log", "a") as f:
f.write(
f"\n\n-------------------------Tortoise TTS Scripts Logs, {datetime.now()}-------------------------\n"
)
main()