Ollama adds local text-to-speech, 100x faster than real time on Apple Silicon (ollama.com)

Ollama has launched a new text-to-speech capability featuring the Phonon model, which boasts impressive performance on Apple Silicon devices. This state-of-the-art text-to-speech engine, developed by Gradium, achieves speeds up to 100 times faster than real-time, making it the quickest option available. On a MacBook Pro with an M5 Max chip, Ollama outperformed Gradium's own runtime by 35%, while maintaining a word error rate comparable to the industry-standard Hugging Face Transformers. The significance of this development for the AI/ML community lies in its potential to enhance applications requiring real-time speech synthesis, such as virtual assistants and accessibility tools. With Phonon supporting multiple languages and voices that can be easily switched via an API, developers can integrate high-quality voice capabilities into their applications with minimal effort. The model's compatibility with the OpenAI SDK streamlines the adoption process, allowing existing users to leverage Ollama's technology seamlessly.