Problem/Motivation

hi, it would be great to have the possibility to use tokens in the ALT field or add the (for example) title field via checkbox on the configuration site. The reason - i run an online webshop for stamps and AI can´t recognise which type of stamp is on the picture - which colour nuance, which perforation, paper type etc., but this exact information is needed for SEO. So the combination of an exact declaration of the product + a visual description of the product would be the perfect fit.

Steps to reproduce

Proposed resolution

Remaining tasks

User interface changes

API changes

Data model changes

Comments

mp created an issue. See original summary.

mp’s picture

mp commented
Issue summary: View changes
w01f’s picture

w01f commented

I think we have a similar basic question.

We have an education resource site where many of the uploaded images do have descriptive titles or a location field on the media item for where the picture was taken.

Is there a way we can reference those other media fields in the prompt so the AI can incorporate existing title details and picture location information into the generated alt text?

sunchaser’s picture

sunchaser commented

Same request here. We offer products with different product names.
can't just have a generic: "woman sitting behind
" it needs to have the actual product name that can be derived from many places like:
- page title
- image file name even
- copy on page
- page url (/products/productname)

That said: thanks already for the effort put into this module. I don't want to sound ungrateful, just trying to make it better :)

edit: I guess we'll have to be able to educate the LLM about our product names tho - cant expect openai to know
is of type

anybody’s picture

anybody
Primary language German
Location Porta Westfalica
commented

I think technically that should be possible using token replacements. Could someone who needs this prepare a MR?