Image generative AIs are superiorly gifted, the matter is clear. However, it often happens that we are dissatisfied with a particular item. And traditionally, it has not been easy to correct it, if we ask it, theAIAI generative most often creates a completely new image.
Dall.e 3 was updated in spring 2024 which now allows you to modify only part of an image. We will see here how to implement this generally unknown functionality.
Image generative AI: how to obtain extraordinary images with your prompts?
Editing tools
Dall.e 3 is accessible from the paid version of ChatGPTChatGPT. It is also possible to access it for free from Copilot from MicrosoftMicrosoft but at the time of writing, this version does not offer the editing tools that we are going to discuss here.
The main point is that, once you have generated any visual with Dall.e 3, you need to click on this image. From there, a few tools appear. The one that interests us is the most on the left, it is a selection tool. When you click on it, you will see a slider intended to influence the size of the selection brush.
Delete part of an image with Dall.e 3
Intelligently erasing part of an image usually requires using a specialized tool, like in this tutorial: How to easily erase elements from a photo. Now Dall.e 3 can do this directly.
Test it on an image. Select an area you would like to delete. It appears covered in blue. Once the selection has been made, type a prompt on the right, for example: “delete this”, “delete this part”… You can be more specific: “remove the cat”, “delete the tables from wallwall »… As you will see, the erasure is carried out while respecting the logic of the image.
Replace part of image
Replacing a part of an image follows the same logic: you must select the area(s) to be replaced then give an appropriate instruction: “replace the helmet with a hat”, “add a necklace to this woman’s neck”… It n does not always execute the exact instruction requested and you sometimes have to reformulate your request until you obtain the desired result. The fact remains that we have here a very useful tool because it overcomes a shortcoming that generative AI has traditionally had and it should lead you to produce images closer to what you specifically want to obtain.
