Speech models in Ollama

Ollama 0.40.11 introduces local text-to-speech on macOS through Phonon, a family of small models from Gradium that can generate 24 kHz mono WAV or PCM audio using command-line tools or an OpenAI-compatible API. Accelerated by Apple’s MLX framework, Ollama reportedly achieved 35% higher throughput than Gradium’s runtime on an M5 Max MacBook Pro while matching the reference implementation’s word-error rate in tests. Users can select from Gradium’s voice catalog, use language-specific models for English, French, German, Spanish, Portuguese, or multilingual speech, stream spoken responses from chat models as they are generated, and save output to WAV files. The base English Phonon model is Apache 2.0 licensed, while the 40-million-parameter language variants use the noncommercial CC BY-NC 4.0 license; Windows and Linux support is planned for the future.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.