Kokoro TTS

Kokoro TTS

Kokoro is a lightweight, fully open-source (Apache-2.0) neural TTS model with 82M parameters that punches well above its weight class in naturalness benchmarks. Unlike many TTS models restricted by non-commercial licenses, Kokoro is free for any use — commercial included.

Key Features

• Apache-2.0 licensed: model weights, code, and training data all permissively licensed
• 82M parameters: runs in real time on CPU; < 200ms latency for short sentences on a laptop
• Multiple voices: 54 built-in voice presets covering US/UK English accents and mixed styles
• High quality: competitive with ElevenLabs on TTS-Arena naturalness benchmarks
• ONNX export: deploy on edge devices without a PyTorch runtime
• Phoneme control: pass IPA phonemes directly for precise pronunciation control

Quick Start

pip install kokoro soundfile

from kokoro import KPipeline
import soundfile as sf

pipeline = KPipeline(lang_code='a')  # 'a' = American English
audio, sample_rate = pipeline(
    "Hello from ai-supply! This is Kokoro TTS.",
    voice='af_heart'
)
sf.write('output.wav', audio, sample_rate)

npx ai-supply add kokoro-tts-lightweight-synthesis

Curated mirror of the open-source Kokoro TTS (Apache-2.0). Get it from the source.