Hi HN — I’m excited to share AICrop ([https://aicrop.app](https://aicrop.app)), a simple tool my collaborator Claude Code and I built to solve a pain point we repeatedly saw among content creators and social media users.
The problem / origin story
Whenever I’m preparing images for social media posts or marketing, I always run into the tedious task of cropping or resizing the same photo to different aspect ratios (Instagram, Twitter, TikTok, LinkedIn, etc.). Doing this manually is time-consuming and error-prone, especially when I want the subject (face, object) to stay centered and look natural across all variants.
Claude Code and I decided: why not make a tool that automates cropping intelligently, but also respects privacy — meaning images never leave the user’s device?
What AICrop does
* You upload a photo (JPG / PNG / WebP, up to 10MB), and AICrop runs object/subject detection entirely in the browser using TensorFlow.js. * It proposes crop frames for major social platforms (Instagram, Twitter, TikTok, LinkedIn, YouTube, etc.) with instant previews. * You can fine-tune the results manually if needed. * No server upload, no account required, no watermark — everything happens locally in your browser.
We built it through “vibe coding” — an iterative, creative collaboration between me and Claude Code. Instead of strict task division, we explored ideas dynamically, letting intuition and quick feedback loops drive progress. It turned out to be both fun and surprisingly productive.
What’s ready now
* Fully working browser version you can try out today (no signup barrier) * Support for the major social media aspect ratios and previews * Manual adjustment and export options * 100% local processing to ensure privacy
What feedback / help we’re looking for
* Does the UX feel smooth and intuitive? * Any crop sizes or platforms we should add next? * Should we build features like batch processing or smarter detection modes? * Thoughts on sustainability — would you keep it free, freemium, or open source? * Any libraries, models, or optimization tips for better in-browser performance?
If you try AICrop, I’d love to hear your thoughts, bug reports, and ideas for improvement. I’ll be around in the comments to chat — thanks for checking it out!