Problems with AI image to image generate feature

Here is a perfect example in which multiple repeats of the AI image to image feature keeps adding changes to the result that were not requested. The first picture is the original image in SculptOK, the second is what Sculpt OK deliver when my instructions were to: “make the old woman’s hair is silvery grey color”. The third picture is what I got on the first try with CHAT GPT.

Can this be fixed by changing coding in the AI image to image feature?

Don’t use the AI optimization in the drop down menu

I agree with you. It doesn’t follow explicit instructions. It takes multiple tries. It changes things when you request that specific things aren’t changed and adds things that you didn’t ask for. I typed in the phrase, “Are you making these mistakes on purpose?” after 8 tries. I used chatgpt to get a proper image and it worked on the first try. Their Ai has minor brain damage. They need to fix the code. I am on one of the $$ tiers. Depth map is fairly acceptable in most cases, but I needed to tune the last map on another platform. This site has lots of bells and whistles. I hope they can improve the AI.

I have switched to using ChatGPT to create images that I want. I have discovered ChatGPT is capable of progressive AI learning. I have told it when I am totally satisfied with a result I enter “Excellent!” and I want it to save what we both learned, i.e. (a) ChatGPT learned how interpret the vocabulary words I had input; (b) I learned to use those input vocabulary words every time I want ChatGPT to create that kind of drawing. Every time I input “Excellent!” at the satisfactory completion of a new chat, ChatGPT stores that new learning. After a few weeks of this ChatGPT gets it right overtime on the firs try. I have also used it to create a factual document succinctly including all the important details from a disorganized list of facts about a topic. I had had it write poetry, limericks and original song lyrics for a music file I want to create. I also have learned when I tell it my input is a list of facts about a topic, it automatically goes to the web and fact checks what I had input. If it finds errors in my factual input, it corrects them and tells me about it.

If you do not use any of the AI options this should not happen. These are in the drop downs. Look for the >

There is a problem with Text that is being worked on. The AI does get confused with creating the text that you want.