Khroma

Khroma is an aesthetic discovery tool. A browser-based color tool that trains a neural network on colors you pick, then generates palettes you can preview as typography, gradients or custom images, filter by hue and hex, and save with hex, RGB, CSS and WCAG contrast details.

Khroma is a browser-based color tool built around a single idea: instead of browsing other people's palettes, you teach an algorithm what you find attractive and it produces combinations tuned to that taste. The training step asks you to pick a set of colors you respond to. That selection feeds a neural network powered algorithm that then generates palettes in your preferred range while suppressing the colors you rejected. The claim is that Khroma has absorbed thousands of popular human-made palettes from across the internet, so the output leans on combinations that already work rather than pure randomness. The generator is where most of the work happens. Palettes can be previewed as typography, as a gradient, as a flat palette, or applied to a custom image, which makes it easier to judge how a scheme behaves on a real layout instead of as isolated swatches. A search layer lets you filter results by hue, tint, value and color, and you can also query by hex or RGB value when you already have a specific shade in mind. That combination of training plus filtering is the practical core of the tool. Useful results go into a personal collection. Khroma keeps an unlimited library of saved combos and, for each pair, surfaces color names, hex codes, RGB values, CSS code and a WCAG accessibility rating. Those details are what make the library usable after the fact: you can pull the CSS straight into a build and check contrast before shipping. The audience is designers and front-end developers who pick colors often and want a faster starting point than a generic palette site. It sits somewhere between a palette browser and a preference-trained generator. Anyone who needs full design tooling, asset export pipelines or team libraries will still finish the job elsewhere, and the tool does not replace a contrast checker for complex accessibility audits. Where it fits best is early exploration and building a reusable, personal color reference.