Build image tags from what pixels can prove

Tags have two very different jobs. Some describe facts in the file: landscape frame, blue, low light, transparent background, busy detail. Others name the subject: bicycle, Jaipur, Labrador. This page does the first group because the pixels can support it. It leaves the second group to someone who can actually see the picture.

That boundary makes the result smaller, but dependable. The copied line is useful as a starting filter in an asset library and as a checklist beside an alt-text field. It is not sold as automatic cataloguing when no recognition model exists behind it.

A small thumbnail is read for luminance, saturation, coarse colour families, edge density, alpha and aspect ratio. Those measurements become human labels. The original file never leaves the tab, and no hidden filename or metadata is used to guess content.

• No object, face, place or brand tags. Those require a vision model, and this build has none.
• No confidence scores dressed up as facts. The photograph-versus-graphic reading is a pixel heuristic and is presented as one.

• Alt textWrite the alt attribute yourself, with the picture open and the findings live.
• CaptionThe visible line under a picture, read back against the job a caption actually has.
• Long descriptionThe paragraph a chart or a diagram needs when one attribute cannot hold it.
• Alt text checkerPaste alt text you already have and read the findings. No picture required.