How Cephable Applies On-Device AI to Accessibility
summary
Cephable runs its entire camera and ML pipeline locally using NPUs and ONNX runtime, converting face expressions, head movements and voice into app controls; a demo shows 1080p at 30+ FPS tracking and mapping gestures to VS Code edits while GitHub Copilot handles text, with measured system load of CPU under 15% and near-zero GPU touch on a Copilot+ PC whose NPU offers ~45 TOPS. Settings sync via Internet only; core inference stays private and offline.
AI-written summary · about
chapters (8)
- 0:00The need for accessibility
- 0:49Introduction to Seable
- 2:58The role of the NPU
- 5:58Efficiency benefits
- 6:43Live software demo
- 9:40Performance comparisons
- 11:20Local privacy and power
- 13:03Getting started with Seable
related talks
57mApplied AI and Accessibility to Play Games in New Ways
55mMaking Every Game and App More Accessible with .NET and AI
20mComputer Vision at the Edge with OpenVINO
36mApplied AI on the edge
36mMaking Technology More Accessible with .NET and AI
31mBringing AI on-Device: From Cloud to Edge
20mScaling Edge AI with Elixir and Nerves
24mMake your app smarter