For a person who is blind or has low vision, a painting is usually reduced to a few words of alt-text, if anything at all. That text rarely conveys the colour, texture, mood, or arrangement that a sighted visitor would naturally take in, leaving the artwork largely inaccessible.
The CANVAS system, described in a recent arXiv paper, proposes a different approach: using AI to generate narrative visual-audio captions. Rather than a bare label, the aim is to offer a richer, more evocative description that captures the qualities of the painting that matter for understanding and appreciation.
The paper highlights the gap between the richness of visual art and the poverty of current text alternatives. By moving toward narrative audio descriptions, CANVAS could make both museum visits and online art collections more meaningful for blind and low-vision audiences.