The world is moving
on-device.
vl8 is an open-source on-device speech SDK with federated learning. Silicon-agnostic across NVIDIA, Qualcomm, Apple, Intel, and embedded ARM. Private beta with design partners.
An open-source speech SDK, silicon-agnostic by design.
On-device ASR
Whisper-class transcription, sub-100ms on mobile SoCs, fully offline. No cloud dependency, no per-inference cost.
TTS + voice agents
Speech synthesis and low-latency conversational AI on-device. Real-time voice interaction without a round-trip to the cloud.
Federated adaptation
Models personalize per-user without uploading audio. Privacy-preserving by architecture, not policy.
Silicon-agnostic runtime
Same SDK across Jetson, Snapdragon, Apple Silicon, Intel, embedded ARM. Our benchmarks don't care which chip wins.
Private beta. Public code shipping soon.
