Observed arrival · 2026-11-07
ribt: a voice in 17.6 milliseconds
An English speech model designed to start speaking quickly, with an interactive browser demo and GPU builds.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Robot builders needing fast English speech
- Worth noticing
- The 17.6 ms first-audio figure is a warm, batch-one M4 Max result; Kokoro is listed at 349 ms.
Field notes
The pipeline starts with lilt, which the page says runs on the CPU in under a millisecond, then assigns phoneme timing and pitch before generating a spectrogram and audio. Its comparison reports 17.6 ms median time to first audio on a warm, batch-one M4 Max, while also noting Kokoro’s slightly higher UTMOS score. The page links to code and weights and says the browser build matches the native build within 6e-5 on the same weights.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue