Ultra-small TTS model Kitten TTS: Only 15 million parameters
The KittenML team released the open-source lightweight text-to-speech model Kitten TTS, which is only 25MB in size and has 15 million parameters. It supports CPU operation without requiring a GPU. The model provides high-quality multiple voice options, optimizes inference speed for real-time synthesis, and comes with an easy installation guide. Currently in developer preview, it will future release full model weights, mobile SDK, and web version, promoting the popularization and application of speech synthesis technology.