VoxRT
A company focused on on-device speech AI. It provides an on-device speech recognition SDK under a hybrid licensing model, with an open wrapper layer and a proprietary runtime.
Timeline (last 90 days)
The VoxRT team built speech AI for the edge (on-device) from the start of the project, rather than migrating from the cloud.
VoxRT's wake word model fits in 100 KB and accepts a fixed trigger phrase (such as 'Hey Assistant'), shipping only one validated model.
VoxRT's tech stack covers iOS, Android, browsers (via WASM), and Linux (Raspberry Pi, Jetson, servers), with a shared Rust runtime that can be compiled to each target platform.
VoxRT's voice activity detector (VAD) runs at 3% CPU on mid-range Android (Snapdragon 662 level).
VoxRT ships the VAD only on mobile, not in the browser, because the WASM configuration differences for continuous audio capture are large enough that the engineering investment has not yet been justified.
All Facts (5)
The VoxRT team built speech AI for the edge (on-device) from the start of the project, rather than migrating from the cloud.
50%UnverifiedVoxRT's wake word model fits in 100 KB and accepts a fixed trigger phrase (such as 'Hey Assistant'), shipping only one validated model.
50%UnverifiedVoxRT's tech stack covers iOS, Android, browsers (via WASM), and Linux (Raspberry Pi, Jetson, servers), with a shared Rust runtime that can be compiled to each target platform.
50%UnverifiedVoxRT's voice activity detector (VAD) runs at 3% CPU on mid-range Android (Snapdragon 662 level).
50%UnverifiedVoxRT ships the VAD only on mobile, not in the browser, because the WASM configuration differences for continuous audio capture are large enough that the engineering investment has not yet been justified.
50%