Summary
A Show HN post describes a new runtime for Gradium's Phonon 31M text-to-speech model, built for browsers using WebGPU. The author says they wanted per-word timings so words could be highlighted as audio plays, so they wrote a second runtime with a different design. In this version, the whole generation loop runs on the GPU through WebGPU kernels, which the post says were written for this project by Claude Opus 5.5. The post reports the model package is 44MB, runs at 115x real time, and can start streaming audio in about 3 ms after initialization.
AI-assisted summary based on the listed source.
What happened
I tried Gradium's Phonon 31M text-to-speech model and was blown away by its quality to size ratio. I needed timings for every word to be able to highlight them, so I wrote a second runtime for it with a different design: the whole generation loop stays on the GPU in WebGPU kernels written specifically for this by...
Why it matters
Developers building browser-based voice features may be able to add speech output and word-level highlighting without a large server-side model. The figures come from the author's own post, so independent benchmarks would be needed before relying on them. Watch for measured results on different hardware and whether the project's code and licensing are published.
What this means for you
Hardware and robotics watchers may want to track whether this becomes a product, benchmark, or deployment signal.
Signal Intelligence
Signal Strength 93%
Technical label SOURCE-BACKED
Public Interest 50
Category ROBOTS & HARDWARE
Reader Depth TECHNICAL
Event context 2 sources
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 73
Practical Impact Score 8
Novelty Interest Score 70
Consequence Score 34
Curiosity Score 52
Shareability Score 58
Event context
Claude gets a source-backed update
Claude has a source-backed update with coverage spanning announcement, how to use.
2 sources
2 angles
ANNOUNCEMENT
HOW TO USE
Why this is here
VQV surfaced this signal because it is recent, relevant to AI Voice, connected to Hacker News Newest.
