🛠️ Phonon TTS on WebGPU, 44MB, 115x real time, with live word timings
Summary
Phonon TTS on WebGPU is a custom runtime for Gradium's Phonon 31M text-to-speech model that runs entirely on the GPU using WebGPU kernels. It achieves fast streaming, a small file size of 44MB, 115x real time performance, and live word timings.
Why it’s interesting
It keeps the entire text-to-speech generation loop on the GPU in WebGPU kernels, achieving 115x real time performance and live word timings from a 44MB model.
Target user
Web developers and users looking for fast client-side text-to-speech generation with word timings.
Source metrics: Points 2 · Comments 0
HN discussion · Project
Source: #HackerNews / Show HN
Summary
Phonon TTS on WebGPU is a custom runtime for Gradium's Phonon 31M text-to-speech model that runs entirely on the GPU using WebGPU kernels. It achieves fast streaming, a small file size of 44MB, 115x real time performance, and live word timings.
Why it’s interesting
It keeps the entire text-to-speech generation loop on the GPU in WebGPU kernels, achieving 115x real time performance and live word timings from a 44MB model.
Target user
Web developers and users looking for fast client-side text-to-speech generation with word timings.
Source metrics: Points 2 · Comments 0
HN discussion · Project
Source: #HackerNews / Show HN