generate.py

from tortoise.api import TextToSpeech
from tortoise.utils.audio import load_voice
import torchaudio

# Create TTS instance
tts = TextToSpeech()

# Load prebuilt voice profile
voice_samples, conditioning_latents = load_voice('train_dotrice')

# Run TTS
output = tts.tts_with_preset(
    text="This is a test from Tortoise Text to Speech running correctly on Python 3.10.",
    voice_samples=voice_samples,
    conditioning_latents=conditioning_latents,
    preset='fast'
)

# Save output using torchaudio directly
torchaudio.save("output.wav", output.squeeze(0).cpu(), 24000)  # 24kHz sample rate
print("✅ Audio saved to output.wav")