Gradium open-sources Phonon: 31M/40M edge TTS models with Rust inference stack for 10+ backends
l-m-z · reddit · 2026-10-21
Gradium released its first open-source offering, Phonon — a collection of small text-to-speech models (31M and 40M parameters) designed to run on-device.
The inference stack is fully open-source, mostly written in Rust, and targets 10+ backends including CPUs, GPUs, and NPUs. Models and code are live on Hugging Face, with the team promising more open-source releases ahead.
Related event: Gradium Open-Sources Phonon: Sub-50M On-Device TTS with 1.7% WER(6 posts)→
More from Infra
- DeepSeek's Cluster Serves 155T Tokens/Day at $38 Revenue per GPU-Hour — zephyr_z9 · 2026-10-21
- Anthropic's 25M daily code checkouts broke git; rebuild cut cost from $90 to $4 — aniketmaurya · 2026-10-21
- Inference engines are becoming composable subsystems, says charles_irl — charles_irl · 2026-10-21
- Musk claims Terafab's single-building output would rank in top 20 country GDPs as it courts Korean chip suppliers — elonmusk · 2026-10-21
- Redditor builds cartoon studio to launch '75 Watt Lab', a CPU-first LLM YouTube channel — EffectiveMedium2683 · 2026-10-21
- Vercel AI Gateway goes zero-fee for BYOK, zero markup on list prices, free ACH payments — cramforce · 2026-10-21