Gradium releases Phonon, open-source on-device Text-to-Speech

Gradium announced Phonon, an open-source on-device text-to-speech system trained on NVIDIA H100 GPUs that generates speech locally without cloud connectivity, emphasizing privacy, availability, and low latency for phones, laptops, embedded devices, and web browsers. Available in 31-million- and 40-million-parameter variants, Phonon supports English, French, German, Spanish, and Portuguese, and Gradium says blind human evaluations ranked it above Kokoro, Magpie, and Pocket TTS despite its smaller size. Its inference stack supports CPU, CUDA for NVIDIA GPUs and Jetson devices, WebGPU, and integrations for Rust, Python, Swift, and WebAssembly, with the company reporting substantially faster GPU performance than CPU inference. The release includes prebuilt voices, speed controls, and API-based voice cloning and voice design, while model weights and inference code are available through Hugging Face and GitHub under the Apache 2.0 license.

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.