Show HN: AICrop – AI-powered local image resizer for social media (aicrop.app)

AICrop is a free, browser-based tool that automatically generates social-media-ready crops from a single upload, using on-device AI so images never leave your machine. Built with TensorFlow.js, it detects people, faces, animals, vehicles and 80+ object types to center subjects across 15+ platform aspect ratios (Instagram, Twitter, LinkedIn, Facebook, YouTube, TikTok, etc.). The app promises <3s average processing, real-time side-by-side previews, manual adjustments, support for JPG/PNG/WebP up to 10MB, and no signup, watermark, or data collection. For the AI/ML community this is a practical example of client-side inference enabling privacy-preserving, low-latency ML experiences without server costs. The smart-crop algorithm claims 95%+ detection accuracy and handles edge cases like large or border subjects, while offering a fallback center-crop and manual tweaking when needed. AICrop showcases how TensorFlow.js models can democratize content tooling for creators and businesses, though trade-offs (model size, mobile performance, and the 10MB upload cap) will determine real-world limits. Overall it’s a useful, privacy-first demo of edge ML applied to an everyday content workflow.