One question kept coming back in our engineering standups: getting a wake-word model to ~100 KB is technically possible, but which trade-offs are you willing to accept?
Wake-word models normally sit at 500 KB to 5 MB. Getting to ~100 KB required accepting three specific trade-offs that we underestimated going in.
Vocabulary is tight. A 100 KB model can hold one wake-phrase well. It cannot generalize to arbitrary phrases the way a bigger model can. If you want the phrase to change per user, or if you want it to work in multiple languages, 100 KB starts to hurt.
Hey Product Hunt
Team at VoxRT here. We wanted a wake-word that runs on a $15 Raspberry Pi, fits in 100 KB, and ships without a network dependency. Nothing off-the-shelf hit all three, so we built our own.
Free tier is "Hey Assistant" on all 4 platforms (iOS, Android, Linux, browser via WASM). Custom brand phrase or language is a paid tier (help@voxrt.com), trained by us and packaged with the same runtime. SDK wrapper is Apache-2.0, native runtime proprietary.
Try the [browser demo]
Happy to talk accuracy, hardware, or licensing.