Voice Cloning using Tortoise-TTS on Apple Silicon

clone_voice.py

# See detailed instructions here:
# https://robertharrison.ca/blog/voice-cloning-tortoise-tts-apple-silicon/

# Import PyTorch and Tortoise.
import torch
import torchaudio
import torch.nn as nn
import torch.nn.functional as F
from tortoise.api import TextToSpeech
from tortoise.utils.audio import load_audio, load_voice, load_voices

# Instantiate a Tortoise TextToSpeech object. It will download all the models used by Tortoise from HuggingFace.
tts = TextToSpeech()

# This is the text that will be spoken. Try with more interesting strings.
text = "Hello. This is your clone speaking."

# Pick a "preset mode" to determine quality. 
# Options: {"ultra_fast", "fast" (default), "standard", "high_quality"}.
preset = "fast"

# This is the voice that will be cloned. 
# Set it to the name of the directory you created in the `tortoise/voices` directory.
voice = "robert"

# Generate speech with the custom voice.
voice_samples, conditioning_latents = load_voice(voice)
gen = tts.tts_with_preset(text, voice_samples=voice_samples, conditioning_latents=conditioning_latents, preset=preset)
torchaudio.save(f'generated-{voice}.wav', gen.squeeze(0).cpu(), 24000)