Type
Chat
Describe your idea
There have been a bunch of improvements since I’ve been here to how we can control the text (for example being able to edit the setting and clothing, and being able to see multiple generated images), but there’s still a confusing sort of nexus of how the text, optional instructions for regeneration, and picturize features connect. I’ll explain the problem better first, then a proposed solution.
My current understanding is that the optional instructions affect the text regeneration and then that new text affects the image generation, but if I try to give directions for what I want the image to be, they’re generally ignored. There is a handy /show feature to describe a desired image, but it really messes with the text of the conversation by shifting toward something that causes the desired picture. The ‘edit image’ feature doesn’t address the problem either, because it doesn’t have access to the chat context. This is a notable hole in how to control the chat from an interface that strives and succeeds at delivering a customizable experience.
My specific proposal is either (1) the user should have the same influence over images as text via an ‘optional instructions’ field when regenerating images*, or (2) a feature for ‘compose a picture’. The functionality is very different from ‘edit image’. Instead of taking an existing image and making modifications, this allows specifying what image the user wants at this moment in the story. The image is generated with the full chat context. There are many chat situations where this would be very helpful. For example, there might be three people in the current scene but you want the picture to show a closeup of one of them. Or you want to show what the Kyn is imagining in their fantasy. Or the image generation is just way different from what you have in mind - this happens at least occasionally to everyone I’m sure.
I spent a bunch of time thinking about what feature would solve this overall confusion before writing this. I hope you’ll strongly consider it because I think it would not only remove some existing frustration, but add a functionality people would have a lot of fun expressing and exploring with.
Thank you,
Ash
*Maybe a separate ‘generate with instructions’ button to avoid making the regenerate button require two clicks.
