Show HN: IconPackGen – Generate Visually Consistent Icon Packs with AI

I’ve been building a small tool called IconPackGen, focused on one specific problem: creating cohesive, visually consistent icon packs using AI.

Most AI image tools can generate nice single images, but producing sets of icons that match each other—style, color logic, geometry—is a different problem. IconPackGen is built specifically for that use case.

What it does

- Generates 9-icon packs based on either: - a general theme (e.g., “minimal line icons”, “retro pixel icons”), or - a reference image to match style.

- Optional: users can provide 9 individual icon descriptions for more control.

- One click generates: - 9 main icons - and optionally 9 variations (slight style changes) - each generation can be followed with subsequent requests

Export formats: - PNG, WEBP, ICO, SVG

The SVGs aren’t traced heuristically — they're created using a second model that vectorizes the PNG icons for cleaner results.

Additional features:

- Consistent illustrations (4 per request) — same idea as icons, but with a separate model tuned for illustrations. - Labels — generate styled text/labels that match your icon theme. - UI mockups — generate small UI component mockups, either from text or a reference image (useful as a reference for icon generation too). - GIF export — turn icons into small animated loops using an animation model.

Why I built it

Designing icon packs manually is tedious and expensive, and most AI tools don't preserve style consistency across sets. I wanted something that made icon-pack creation near-instant, especially for indie devs, hobby projects, and internal tools.

I’d love to get feedback from the HN community: - Does the workflow make sense? - What formats or features are missing? - Is this useful for your projects?

Try it out here: https://iconpackgen.com Happy to answer questions or dive into technical details about the models or vectorization pipeline.