Spaces:
Running
Running
Feature Suggestion: Rank results for AI feedback/training
#1152
by
mmroberts
- opened
I don't know how feasible this is with the current Dall-e codebase, but dalle-mini provides an excellent opportunity to fine-tune the AI by allowing users to rank the 3 best renderings from the results list. This information could then be used to give weight to the results that best matched user expectations.
Mockup:
I've been thinking about this too, I think it would really help get the images people are looking for while also being a tool to get the AI to do almost realistic-looking images without having to add descriptors