Reported by 2 outlets
The story so far
AI summary of the reporting below · Updated · How we work
Ollama has released version 0.40.11 of its local AI tool, introducing built-in text-to-speech generation for Apple Silicon Macs. The update incorporates Phonon, a family of compact speech models developed by French startup Gradium. Users can generate voice audio across five languages—English, German, French, Spanish, and Portuguese—via individual models or a unified multilingual version. [1][2]
Built on Apple's MLX machine learning framework, the feature outputs 24 kHz mono WAV files and streams speech in real time, playing audio after the first sentence while text is being generated. Ollama claims its MLX integration runs 35 percent faster on an M5 Max chip than Gradium's native runtime. Developers can hook into an OpenAI-compatible endpoint, though commercial deployment faces a hurdle: while the English model uses an Apache 2.0 license, the other language models are restricted to non-commercial use under CC BY-NC 4.0. [1][2]
The update streamlines entirely offline, voice-enabled AI setups that previously demanded complex custom configurations. While speech support is currently exclusive to macOS, Ollama says versions for Windows and Linux are planned for future release. [1][2]
Ollama, the popular app for running AI models locally, announced that its Mac app now supports text-to-speech generation, with Windows and Linux support coming soon. Here are the details.
via 9to5Mac
Ollama can now convert text to speech entirely locally on Macs with Apple Silicon. Gradium's Phonon models speak five languages, including German, read chat responses aloud as they are being generated, and can be accessed through an OpenAI-compatible interface. There is a catch for commercial projects, however.
via Notebookcheck