Theme Check
Find out whether a report theme is readable, accessible and safe for colour-blind readers, and get a corrected version in one click. Source on GitHub (MIT).
Drop a Power BI theme, a .thmx or a PowerPoint, Excel or Word file
or . Accepts .json, .thmx, .pptx, .potx, .xlsx, .xltx, .docx, .dotx
No file at hand? Try an example:
Your file stays in your browser. Nothing is uploaded, nothing is stored.
Builds a corrected theme you can download
The palette series colours in order, with their contrast on the background
The checks each with what it found and what to change
As colour-blind readers see it ΔE is the CIEDE2000 distance: under 10 looks alike, under 5 looks identical
On a chart your theme on a small report, switch the vision to compare
Why colour blindness is the check that matters most. About 1 man in 12 and 1 woman in 200 have a colour-vision deficiency, almost always of the red-green kind (the deutan and protan types). In a room of 25 people reading your dashboard, expect one or two who cannot tell your green series from your red one. The tool simulates each deficiency with the Machado, Oliveira and Fernandes (2009) model and measures how far apart your colours stay with CIEDE2000, the colour-difference formula the print industry uses. Two series under ΔE 10 get confused in a legend; under 5 they look like one colour. It checks the first 8 series, the ones most charts actually reach.
The numbers behind each check. Text follows WCAG 2.2: 4.5:1 for normal text (success criterion 1.4.3), 3:1 for large text, meaning 18 pt, or 14 pt bold. Chart marks (bars, lines, markers, legend keys) need 3:1 against their background (1.4.11, non-text contrast); a pastel on white sits around 1.5:1 and a thin line in that colour vanishes on a projector. A Power BI theme with fewer than 6 data colours repeats colour 1 on the next series; more than 12 is past what any reader can match to a legend. Good and bad are tested under red-green deficiency because KPI cards and conditional formatting lean on exactly that pair. Fonts are checked against the ones Windows, Office and the Power BI service ship with: a Google font in a theme looks right on the designer's PC and falls back to something else on everyone else's. Dark themes are detected from the background luminance, and the fixes lighten instead of darken.
What "Fix it for me" does. It darkens (or, on a dark theme, lightens) text until it reaches 7:1 for main text and 4.5:1 for secondary text, steps each faint series colour along its own hue until it reaches 3:1, and pulls colliding series apart: first with a lighter or darker version of the same colour, and only if that is not enough with a hue from the Okabe and Ito colour-blind safe set. It removes repeats, adds colours up to 6, picks a good and bad pair that survives red-green colour blindness, swaps uninstalled fonts for Segoe UI, and fills every missing role. You get a Power BI theme and an Office theme. To adjust it by hand, open it in the Theme Builder; for ready-made designs, see the templates.
Privacy and what it reads. The file never leaves your browser: the page unzips Office files and parses JSON locally, with no server behind it. From a Power BI theme it reads dataColors, background, foreground and the other root colours, textClasses and whether visualStyles exists. From an Office file (.thmx, or the theme inside a .pptx, .potx, .xlsx or .docx) it reads the 12 colour slots and the heading and body fonts. It reports a missing role as missing; it never grades a default the file did not set.