Your daily frequent11y newsletter, brought to you by @ChrisBAshton:

AI-Generated Images from AI-Generated Alt Text

Adrian Roselli explores the topic of AI and accessibility, as claims are being made that AI could make up for things like a lack of human-provided alt text. He feeds several different images into the following browsers/software and has them generate alt text descriptions for the images:

• Microsoft Edge
• Google Chrome
• Apple iOS VoiceOver Recognition
• Microsoft Office

He then feeds the generated alt text into AI image generators, Craiyon and Midjourney, to see what they create from the alt text. He does the same with his own manually created alt text, to compare the image outputs.

This is an exercise to highlight that automated alt text still isn’t particularly good, and that the quality of alt text greatly influences the quality of generated images (a metaphor for how a screen reader user uses the web and relies on good alt text to be able to properly understand an image).

There’s a follow-on article, AI-Generated Images from AI-Generated Prompts, which explores the idea of feeding images into tools that then spout out AI prompts you can use to try to generate new images (very cyclical!).

Also cool to note that the opening image of each article was generated by AI, using the title of the post as the seed phrase.

Prefer longer newsletters? You can subscribe to week11y, fortnight11y or even month11y updates! Every newsletter gets the same content; it is your choice to have short, regular emails or longer, less frequent ones. Curated with ♥ by developer @ChrisBAshton.