How do you extract a color palette from an image?
This tool draws your image into a small hidden canvas, reads every pixel of that sample, and runs k-means clustering to group them into six color families. Each swatch is the average color of one cluster, and the list is ordered by how much of the picture each cluster covers, so the palette reflects real area instead of a few loud pixels in one corner. Sampling small is deliberate: a 160 pixel sample holds the color story of a 4K image and clusters in a few milliseconds.
How many colors should a thumbnail use?
Three do most of the work. One background, one color that separates the subject from that background, and one accent for the title. Six extracted colors are a menu, not a shopping list. That is what the five-role row is for: it picks the three that matter and adds two neutrals for outlines and text. If you are working from a thumbnail you admire, the palette is the part worth taking, and the walkthrough on how to recreate a YouTube thumbnail style covers what else to copy and what to leave behind.
What contrast ratio does thumbnail text need?
WCAG asks for 4.5 to 1 for normal text and 3 to 1 for large text. A thumbnail title is large type on your monitor, but in the YouTube sidebar it renders at about 168 pixels wide and stops being large. Treat 3 to 1 as the floor and 4.5 to 1 as the target. If a pairing you love lands under 3 to 1, you do not have to abandon it. Put a thick dark outline on the type and the contrast that matters becomes text against outline, not text against background.
Is my image uploaded when I extract its palette?
No. The file is decoded into a canvas inside your own browser and the clustering runs on your own processor. Nothing is sent anywhere, there is no account and no queue, and closing the tab erases the image and the palette with it. That is also why the tool stays free: your machine does all of the work.