For CLI use case. In case someone wants to use it without gradio

[ProjCRys]

cli.py

import platform
import argparse
import torch
import random
import json
from pathlib import Path
from TTS.api import TTS
import uuid
import html
import soundfile as sf

def is_mac_os():
    return platform.system() == 'Darwin'

def get_available_voices():
    this_dir = str(Path(__file__).parent.resolve())
    return sorted([voice.name for voice in Path(f"{this_dir}/targets").glob("*.wav")])

def random_sentence():
    with open(Path("harvard_sentences.txt")) as f:
        return random.choice(list(f))

def gen_voice(string, spk, speed, english, model_name, file_name="output.wav"):
    string = html.unescape(string)
    short_uuid = str(uuid.uuid4())[:8]
    output_file = Path(file_name)
    this_dir = str(Path(__file__).parent.resolve())
    tts = TTS(model_name=model_name).to(device)
    tts.tts_to_file(
        text=string,
        speed=speed,
        file_path=output_file,
        speaker_wav=[f"{this_dir}/targets/" + spk],
        language=languages[english]
    )
    return output_file

def update_speakers():
    updated_speakers = {p.stem: str(p) for p in list(Path('targets').iterdir())}
    return list(updated_speakers.keys())

def handle_recorded_audio(audio_data, speaker_dropdown, filename="user_entered"):
    if not audio_data:
        return speaker_dropdown

    sample_rate, audio_content = audio_data

    save_path = f"targets/{filename}.wav"

    sf.write(save_path, audio_content, sample_rate)

    new_speakers = update_speakers()

    updated_dropdown = gr.Dropdown(choices=new_speakers, value=filename)
    return updated_dropdown

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Text-to-Speech CLI Application")
    parser.add_argument("--text", help="Text to convert to speech", required=True)
    parser.add_argument("--speaker", help="Select speaker", choices=get_available_voices(), default="Rogger")
    parser.add_argument("--speed", type=float, help="Speed of speech", default=0.8)
    parser.add_argument("--language", help="Language of text", default="English")
    parser.add_argument("--model_name", help="TTS model name", default="tts_models/multilingual/multi-dataset/xtts_v2")
    parser.add_argument("--filename", help="Filename for recorded audio", default="output.wav")

    args = parser.parse_args()

    # Load language data
    with open(Path('languages.json'), encoding='utf8') as f:
        languages = json.load(f)

    # Set device
    device = torch.device('cpu') if is_mac_os() else torch.device('cuda:0')

    # Generate voice
    output_file = gen_voice(args.text, args.speaker, args.speed, args.language, args.model_name, args.filename)

    print(f"Voice generated and saved to: {output_file}")

run-cli.bat

@echo off
python -m venv venv
call venv\Scripts\activate
if errorlevel 1 goto :error

python cli.py --text "Testing testing" --speaker user_entered.wav --language English --filename "output.wav"
pause
exit /b 0

:error
echo Failed to activate the virtual environment.
pause
exit /b 1

To run the cli.py script with command-line arguments, you can use the following examples:

1. Basic example with required arguments:

   python cli.py --text "Hello, this is a test."

2. Example with additional optional arguments:

   python cli.py --text "Hello, this is a test." --speaker Rogger.wav --speed 1.0 --language English --model_name "tts_models/multilingual/multi-dataset/xtts_v2" --filename "custom_output.wav"

python cli.py --text "Hello, this is a test." --speaker user_entered.wav --speed 1.0 --language English --model_name "tts_models/multilingual/multi-dataset/xtts_v2" --filename "custom_output.wav"

user_entered.wav and Rogger.wav is at the targets folder for context.

Make sure to replace the values with your specific requirements. Note that the --speaker argument should be one of the available voices returned by the get_available_voices() function.

[BoltzmannEntropy]

Thanks.