Open-source State-of-the-art TTS model which runs on a CPU π»
Python
"""KittenTTS 2 β get each chunk as soon as it is generated."""
import numpy as np
import soundfile as sf
from kittenml import KittenTTS
m = KittenTTS("KittenML/kitten-tts-2")
text = ("One day, a little girl named Lily found a needle in her room. "
"She knew it was difficult to play with it because it was sharp. "
"Lily wanted to use the needle to sew a button on her shirt. "
"She asked her mom for help, and her mom smiled and said yes. "
"They sat together by the window, where the afternoon light was best, "
"and her mom showed her how to hold the fabric steady with one hand.")
# Each chunk is a 1-D float32 array. Text under ~380 characters yields once.
chunks = []
for chunk in m.generate_stream(text, voice="Luna"):
print(f"{len(chunk) / m.sample_rate:.1f}s")
chunks.append(chunk)
sf.write("output_streaming.wav", np.concatenate(chunks), m.sample_rate)