Tools
kyutai-labs/pocket-tts v2.1.0: 100M-Parameter TTS Runs 6x Real-Time on CPU at 4.2K Stars
Kyutai Labs released Pocket TTS v2.1.0 on May 4 — a 100M-parameter text-to-speech model designed to run on CPUs without GPU dependency. It generates audio 6x faster than real-time on a MacBook Air M4 with ~200ms latency to first audio chunk via streaming. Supports 6 languages (English, French, German, Portuguese, Italian, Spanish), voice cloning from custom audio files, and ships with Python API, CLI, and HTTP server interfaces. 4.2K stars across 8 releases.
Source
↳ Follow the thread