Speech models in Ollama
Ollama now supports text-to-speech, starting with Gradium's Phonon model, on macOS for now with Windows and Linux planned. It works through the OpenAI-compatible /v1/audio/speech endpoint and requires Ollama 0.40.11 or later.
What the article says
Ollama can now turn text into speech, according to a post dated October 24, 2026. The first speech model is Phonon, a family of small text-to-speech models from Gradium. For now the feature runs on Apple Silicon Macs; Windows and Linux are planned but not yet released. Users need Ollama 0.40.11 or later, then run ollama pull phonon.
Ollama relies on Apple's MLX framework and says it is 35% faster than Gradium's own runtime on a MacBook Pro with M5 Max, with a word error rate matching the reference implementation. The English model is under Apache 2.0, while the 40M models (French, German, Spanish, Portuguese, multilingual) use CC BY-NC 4.0, meaning noncommercial use only. Phonon works with the OpenAI-compatible /v1/audio/speech endpoint, streaming 24 kHz mono audio.
Read the official article AnnuIA profile
Summary written by AnnuIA in its own words from the official page of Ollama. The content and rights belong to the publisher: please read the original article.