Local & Real-Time Tools: Nexa SDK and Sokuji

Two open-source tools address local/offline AI needs and real-time speech translation: Nexa SDK for on-device ML and Sokuji for live-in-meeting translation.
Key insights
Sokuji enables real-time speech translation across platforms
Sokuji is an open-source real-time speech translation tool integrating OpenAI and Gemini models to capture microphone input, translate and play back audio. It offers desktop apps and browser plugins for Chrome/Edge, adapts to Google Meet and Teams, creates virtual audio devices on Linux, and supports real-time audio passthrough and visualization.
Nexa SDK offers full local offline AI stack
Nexa SDK is an open-source toolkit for on-device models emphasizing model compression and local search across thousands of images and documents milliseconds. It supports offline operation for privacy, scalable local document handling, CLI tooling for model download and run, and compatibility with OpenAI API semantics for integration.