April 29, 2026
Palette drama in living color
Creating a Color Palette from an Image
Internet swoons as a photo color picker turns messy pixels into shockingly good vibes
TLDR: A developer rebuilt their photo-to-palette tool from scratch to make color picks feel more human, and readers were seriously wowed by the results. Most comments were glowing, but a few sparked debate over edge cases and whether AI will eventually bulldoze handcrafted methods like this.
A nerdy blog post about pulling five colors out of a photo somehow turned into a full-on comment-section lovefest. The author basically confessed their first attempt was a chaotic monster: too many fixed rules, too many weird exceptions, and not enough trust that the results would look good on fresh images. So they scrapped it, rebuilt the whole thing with a cleaner approach, and chased a very human goal: show the facts of the image’s colors, but keep the feeling too. In plain English, they wanted a palette that looks like something a person would actually choose, not something a robot coughed up.
And the crowd? Wildly impressed. One commenter flat-out declared it might be the best color palette generator they’ve ever seen, which is the kind of praise that makes the rest of the internet sit up straight. Another dropped a string of rival and related palette tools, giving the thread that classic tech-forum energy: applause mixed with a mini “you should also check out these other wizards” showcase. There was even a thoughtful nitpick about “phantom blue” shadows, with one user wondering whether the tool should care about where colors appear in the photo, not just what colors exist.
Then came the spiciest subplot: AI creeping into the chat. One commenter said newer image models may be teaching everyone the “bitter lesson” even in color picking, hinting that handcrafted tricks could get steamrolled. So yes, the palettes are pretty — but the real drama is whether careful human-made logic still has a future when the machine-learning crowd is knocking at the door.
Key Points
- •The article replaces an earlier RGB/HSL heuristic palette generator with a new OKLCH-based clustering system.
- •The new method uses K-means++ to find candidate color regions, then merges or rescues clusters to arrive at five palette colors.
- •OKLCH chroma is presented as a more reliable indicator of colorfulness than HSL saturation, particularly for very dark pixels.
- •Cluster merging initially uses a distance threshold of 0.07, while a rescue pass can add missed colors if they are distinct and cover at least 0.1% of the image.
- •Testing on twelve sample images led the author to prefer K=14 and to weight chromatic differences twice as much as lightness differences during merges.