blib-la
Multiple gpt-vision guidelines
As a user, I want to use different guidelines for gpt-vision, so that I can choose from a list to cover my use case
I expect to have a collection of guidelines * example responses. I want to be able to easily switch between them It would be nice if I could also share them with a friend
@MNeMoNiCuZ Guidelines tell GPT how to write your caption.
We offer two options: guidelines, example response.
These options help guide the AI to create consistent captions.
e.g.
Tags:
Guideline: Please caption the image using booru-style tags
Example Response: ["1girl, red skirt, school uniform", "1boy, military uniform, looking at viewer"]
Storytelling:
Guideline: Please caption the image in a story telling style
Example Response: ["the girl is walking down a road in a deserted city, the dystopian landscape filled with dusty air as dark clouds cover the burning sky", "a rotten apple came to life, eyes appear on its skin and arms and feet grow as it lifs itself from the bowl wehre other fruits remain lifeless"]
These can be very precise (this is one from my latest model-preparation)
Ah, great. Yeah this should certainly be a list of customizable instructions.