Show HN: I run a vision model on every screenshot, locally, on a 4GB GPU (github.com)

A new open-source project called ScreenMind has been unveiled, offering a privacy-first alternative to existing screen-aware AI tools by utilizing Gemma 4's multimodal capabilities. ScreenMind captures your screen activity and analyzes every screenshot locally, ensuring that sensitive data stays on your device without reliance on cloud infrastructure. Features like smart capture, semantic search, voice memo transcription, and automatic meeting summaries enable users to build a comprehensive AI memory that can be queried conversationally. This development is significant for the AI/ML community as it addresses growing concerns over privacy and data security with tools like Microsoft’s Recall, which faced backlash for storing data in plaintext and sending telemetry. ScreenMind promotes a fully local operation with robust features such as real-time emotion detection, adaptable analysis modes, and a full agent/plugin system for automating tasks. By leveraging advanced techniques such as EasyOCR for text extraction, spatial layout recognition, and hybrid search algorithms, ScreenMind provides an efficient and user-friendly approach to managing digital experiences, all while safeguarding user privacy.