Mozilla Firefox 130 will introduce an experimental new capability to automatically generate alt-text for images using a fully private on-device AI model. The feature will be available as part of Firefox’s built-in PDF editor, and Mozilla’s end goal is to make it available in general browsing for users with screen readers.
For a preview and early feedback check Firefox’ Nightly wiki page.
Futher reading and source:
Tarek Ziadé, Experimenting with local alt text generation in Firefox Nightly (blog post)
I like that the model can be ran locally and it seems useful to me. I am also concerned that it might seem as if people don’t need to write alt-text, since an AI model can generate it.
I supposed I’d rather have AI generated alt text than no alt text but i agree with Jan, I appreciate when people write really good, descriptive but also contextual alt text. Sometimes there’s some nuanced cultural meaning in an image that I suspect AI won’t pick up until it’s trained better.
A different cultural meaning is indeed a good point, that I haven’t thought of yet. I think, purely factually thinking developers or a blind AI might also just neutrally describe the content of the image, e.g. „A little, yellow figure of a cat with one arm lifted“ (Western culture) vs. „Golden Maneki-Neko with inviting gesture“ (Chinese culture)
An interaction designer once said: in IxD, context is king. GenAI also needs context to return good results. Therefore, giving the gen. AI some context would be a good idea. IMHO it could be provided from the website or page itself or user preferences in the browser, such as text content or language settings.