Ollama adds local text-to-speech support for Mac

Ollama, an open-source platform for running AI models locally, has announced that its macOS application now supports text-to-speech (TTS) generation. This update allows users to convert text into spoken audio without relying on cloud services, enhancing privacy and enabling offline use.

The initial speech capability utilizes the Phonon model family developed by Gradium. On macOS, Ollama leverages Apple’s MLX machine learning framework to utilize the unified memory architecture of Apple Silicon. In testing on a MacBook Pro equipped with an M5 Max chip, the platform achieved speech generation speeds 35% faster than Gradium’s own implementation while maintaining comparable accuracy.

The new feature supports real-time speech generation, allowing users to hear responses as they are being produced. Currently, Phonon offers models for English, French, German, Spanish, and Portuguese, including a multilingual model and various accent options. Support for Windows and Linux is expected to follow.