KittenML releases open-source 1.58-bit TTS model Kitten-TTS 2

KittenML open-sourced the code and weights of Kitten-TTS 2, calling it the world's first 1.58-bit ternary-weight TTS model. The 1.7B-parameter model runs in real time on CPU alone, supports zero-shot voice cloning from under 5 seconds of audio, 20+ languages, and emotion/emphasis control, offering an open-source alternative to ElevenLabs for GPU-poor users.

• ElevenLabs launches two real-time speech-to-speech models2026-10-20
• A16z report: consumer AI opportunities and untouched categories2026-10-09
• Customer workflows emerge as the AI era's new moat2026-10-09
• ElevenLabs doubles valuation to $22B in employee tender2026-09-30
• ElevenLabs launches v4 speech model with more expressive, consistent voices2026-09-29