Is it possible that in some way by using a image action that it retains the current picture and just changes a minor detail? SO far Nano Banana and GPT Imagen 1.5 always generate new pictures… Am I misisng something? Is it a prompt problem?
Hi, could you share a more specific scenario?
I’ve used Pickaxe for things like adding a hat to a person or changing the color of objects in images, and it works decently…it’s not always reliable, which is more of a general AI limitation. 
Even when I try the same prompt on Gemini or ChatGPT directly…sometimes it follows instructions well, and other times it gives a completely random result.
Hoping image editing continues to improve. on similar note…i am also really excited for video editing once it becomes more accessible/usable.
I have the AI crete Illustrations with Vocabulary written on it… it works nice but 9/10 times it has one word twice … I would like to be able to tell it just earse this text label once. In Gemini I can do this and it works… I am using Nano Banana 2 for the image generator …
I was able to replicate the issue with Nathaniel on the office hours last week. Pickaxe can edit uploaded images but can’t edit generated images. So if you generate a photo of a man, and then prompt it to give him a hat. It will generate a new man and give him a hat rather than putting a hat on the original man.
working on this btw ! ^
1 Like