Kitten TTS — Lightweight Browser Voice
8 expression-based voices in a tiny 24MB model. WebGPU + WASM, configurable sample rate, 100% offline.
About Kitten TTS
Kitten TTS is the lightest engine available on OfflineTTS. At just 24MB, it downloads in seconds and runs on virtually any device — from desktop browsers to mobile phones and embedded hardware.
The 8 expression-based voices cover a range of tones: cheerful, serious, sad, whisper, excited, gentle, calm, and neutral. Each voice is a compact embedding that shapes the model's output character.
Kitten TTS supports configurable sample rates from 8kHz to 48kHz, making it versatile for different quality and performance needs. WebGPU acceleration is available on supported browsers for faster generation.
Compare engines: Kokoro TTS (54 voices, 9 languages, highest quality) · Piper TTS (25 voices, fastest on CPU)