The new integration between Ollama and Phonon models enables text-to-speech conversion using natural-sounding AI voices on Mac devices, without relying on cloud services or paying subscription fees.
اضافة اعلان
Apple’s App Store has seen widespread adoption of speech-to-text applications, making voice-based typing on Mac devices easier than ever. One such application, TypeWhisper, was previously covered.
However, high-quality text-to-speech models remain limited, and most of the good options require paid subscriptions.
Ollama aims to change this. It is an open-source platform that allows AI models to run locally on computers. It supports macOS, Windows and Linux without storing users’ text prompts or using them to train models.
How is Ollama changing the text-to-speech experience?
Ollama now supports running text-to-speech models on Mac devices, starting with Phonon models developed by Gradium.
Phonon includes a collection of small AI models capable of converting any text into speech directly on a Mac, without sending data to external servers. Support for Windows and Linux is expected to arrive soon.
The voices showcased by Ollama on its blog sound highly natural. Users can choose from several voices, including Zoey, Easton, Freya and Hugh, with American and British accents.
The models are not limited to English; versions supporting French, German, Spanish and Portuguese are also available.
In terms of performance, speed is not expected to be a major obstacle for users.
Thanks to Apple’s MLX framework, Ollama says Phonon models run 35% faster than Gradium’s runtime environment on a MacBook Pro equipped with an M5 Max chip, without affecting output accuracy.
How can the new feature be used?
Users can connect Phonon models to one of the AI models designed for conversation, allowing text responses to be read aloud.
Audio playback begins as soon as the first sentence has been generated, meaning users do not have to wait for the entire response to be completed before listening.
To use this feature, users need a Mac equipped with an Apple Silicon chip and Ollama version 0.40.11 or later.
They can then download a Phonon model through the platform to begin using text-to-speech locally.
It is worth noting that the English-language model is available for free under the Apache 2.0 license. However, the models designed for French, German, Spanish and Portuguese, along with the multilingual models, are restricted to non-commercial use.
