AI Image: Custom - prompts not copying/carrying over

Losing the text prompts is really annoying. I used to be able to go into other people's courses and see what prompts they'd used and then use in my courses. Now everything is lost when published and we'll have to keep our own copies locally. This is a backward step.

The iterations feature is good but often I ask it to do something, it says it has done it, I look at the image and nothing has changed. I'd be quicker redrawing the image myself as its ignoring me.

Example: Make all the boxes the same size > no change > they are not all the same size> The image now shows a beautiful sunset over the ocean. 

Example: Editorial Sketch > The image now features a lively cityscape at dusk. > Photorealistic > Here's a stunning landscape with vibrant colors.

Also, the more edits you do the more scrolling you need to get back to preview the image on the right side.

I agree, we need the old image generator back while this is refined.

I agree that the image generation has taken a step backward; I noticed it especially in terms of spelling.  When I first started using it last spring it couldn't spell anything correctly.  Over time, it became really good at spelling, and I was very pleased.  But now it's back to not even using actual letters, just random squiggles.  And of course it is labeling things I did not ask it to label in the first place, making the errors unnecessary as well as annoying.  I did find, however, that I could give it an image I wanted to incorporate into the picture, and it used it.  I hope that is predictable.

Here's an example. I had created this image using Vector Blues non-legacy with the prompt >>Building with a sign "Animal Shelter" on it and a group of people of different races and ages and one person is holding a dog on a leash and the building and the group of people are separated by a brick wall<<

To try to re-create it using the new image generator, I tried many prompts. They all came out with about the same image, and this is the closest I could get >>A cartoony image of a building with a sign on it that reads "Animal-Shelter" and a group of people of different races and ages and one person is holding a dog on a leash and the group of people and the animal shelter are separated by a tall brick wall to indicate that the shelter is not accessible to those people<<

This is not only too AI-looking, but the generator introduces elements that our learners will recognize as not like an animal shelter. I have tried MANY iterations and the style is always the same and each time the generator adds details I don't want. 

If anyone has suggestions, I would appreciate it. Or if Articulate wants to return to their old system while they refine the new one, that works. :)

I'm having the same problem and came on here to find advice! I started a course with flat, 2-D images, using the Vector Blues choice (sometimes legacy and sometimes non-legacy). I can't get any similar affect with any prompts I have tried with the new AI image generator. The characters and scenes look too AI, if that makes sense. Just now I kept trying to get the image flatter and the image generator kept making the female curvy, in tight jeans. So, the same problem with lack of variety in output but also not the similar tone carrying through that we had when we could choose a style. 

Does anyone have suggestions for image prompts that would produce a style like the new or old Vector Blues? I've tried the phrase Vector Blues, I've tried flat, 2-d, cartoony...I feel like I've tried everything and get the same style every time.