Ollama, the open-source platform for running AI models directly on your own machine, announced that its Mac app now supports local text-to-speech generation. The first speech model available is Phonon, a family of small text-to-speech models built by Gradium. According to 9to5Mac, this is the first time Ollama has supported speech generation at all.
The feature is live now on macOS, and Ollama says Windows and Linux support is coming soon. Because everything runs on the device, audio is generated without an internet connection and without sending text to external servers — a meaningful shift for anyone using Ollama for private or offline work.
Ollama local text-to-speech: what Phonon can do
Ollama says speech is produced in real time, so you can start listening to an AI-generated answer while it is still being written rather than waiting for the full response. Users can also pick from several voices with different accents, depending on which model they download.
Notebookcheck reports that the models are reachable through an OpenAI-compatible interface, which makes it easier to plug them into existing tools and workflows. The same report flags a catch for commercial projects: licensing terms for business use are worth checking before shipping anything built on Phonon.
- Speech models run locally on Apple Silicon Macs through Ollama 0.40.11 or later
- Phonon supports five languages: English, French, German, Spanish and Portuguese
- Separate models per language plus one multilingual model covering all five
