The open-source platform Ollama has announced that its Mac application now officially supports text-to-speech generation, with Windows and Linux versions to follow soon. This marks Ollama's first introduction of voice generation capabilities, signifying a significant step forward for local AI applications toward voice output. The first supported voice model is the Phonon series developed by Gradium, which features a compact size and advanced technical specifications. Users can download and run it directly on Apple Silicon Macs without relying on cloud services, making it particularly suitable for privacy-focused and offline usage scenarios. On macOS, Ollama leverages Apple's open-source machine learning framework MLX, combined with the unified memory architecture of Apple Silicon, significantly enhancing operational efficiency. Official tests show that on a MacBook Pro equipped with an M5 Max chip, the model achieves a voice generation speed 35% faster than Gradium's own implementation while maintaining a comparable word error rate to the reference implementation. The new feature also supports real-time voice generation, allowing users to hear the voice output synchronously during the answer generation process without waiting for the AI to complete the full text output, making the interaction experience closer to natural conversation. Phonon currently offers standalone models for five languages: English, French, German, Spanish, and Portuguese, as well as a multilingual model supporting all five languages. Users can also choose from different accents, such as American or British English. Ollama is a popular open-source AI project that allows users to download and run open AI models locally or access more powerful models in the cloud as needed.