Visual asset generator

Cate it needed a clear visual identity to establish trust in a marketplace context—where the sender of every communication needs to be immediately recognizable. I'd been curating a moodboard on Are.na with visual directions that could work for the brand, but my illustration skills are limited.

Rather than following the traditional approach, I saw an opportunity to experiment with AI image generation models to create branded visuals, social media content, and illustrations by mixing references from my moodboard.

I started on Replit, an app builder that handles both front-end and back-end infrastructure. I've worked with similar tools before (like Webflow for my previous portfolio), but Replit's strength is integrations—which made connecting to my Are.na moodboard straightforward. For image generation, I integrated Google's models through Replit's built-in AI integration, which simplified billing.

With the integrations running, I focused on the interface. I've been interested in node-based interactions lately, particularly for AI tools—I like working with visual references, and nodes let me connect and compare multiple variations easily. I used React Flow to build the interface on Replit.

The tool worked, but I didn't want to keep paying $25/month for Replit hosting. I used Claude Code to restructure the architecture and migrate the app to my own infrastructure—my own database, my own Google AI Studio billing, my own hosting. I wanted to own the tool, not depend on a platform.