Drop in a published SCORM package or a PowerPoint deck and see every image at once, with the description a screen reader would actually announce. It finds the missing ones, the filenames pasted into the field, and the placeholder text that reaches production because it satisfies every check that only asks whether something is there.
Free · No account · Your files never upload
A published SCORM or web package, or a .pptx. It is read on your computer and never uploaded.
Useful for a single page or a snippet from a web object. Dropping the published package checks every page at once.
Opens a published course and lists every image with the description a screen reader would announce for it. One pass over the whole package, rather than clicking through forty slides with the accessibility pane open.
An image with an empty description is marked decorative, and screen readers skip it. That is correct and it is common: dividers, background flourishes, an icon sitting beside a label that already says the same word. Tools that flag every empty description teach people to write “decorative image” into a spacer, which is worse than silence, because now it gets read aloud. This never reports an empty description as a problem. What it does report is the difference between empty and absent, which is the distinction most checkers collapse: an empty description is a decision, and a missing one is an oversight.
Alt text that exists and is useless. A filename, a default shape name like “Picture 3”, or the word TBD. Any tool that asks only whether a description is present will pass all three, and a course can reach one hundred percent coverage while telling a blind learner “safety_diagram_v2.png”. This is not hypothetical: export tooling routinely fills the description field with the filename when the author left it blank, so the fault is manufactured rather than lazy.
Because that is what an auditor evaluates and what a learner receives. A project file holds the author’s intent; the published package holds what assistive technology will actually meet, including markup contributed by states, layers and web objects that the project view never shows you. A Storyline .story project will open here and be refused for that reason. Reporting a clean result on a course full of undescribed images would be the worst outcome this tool could produce, so it declines rather than guesses.
It cannot tell you whether a description is accurate. Judging whether “a worker checking a gauge” describes the image in front of you requires seeing the image, and no automated check gets there. It also has no view on whether an image should have been decorative in the first place. Treat a clean result as the floor, not as a conformance statement.