You’ve no doubt seen the plethora of AI art generators that have emerged in the last year or so: super-smart engines that can produce images that look exactly like real photos, or art created by real humans. Work. As time goes on, they’re becoming increasingly powerful, and adding more and more features—you can even find an AI art tool in Microsoft Paint now.
New to the DALL-E AI image model, available to ChatGPT Plus members paying $20 per month, this capability Edit parts of the image.Just like you can in Photoshop: you no longer have to create a whole new image just because you want to change one element of it—you can show DALL-E the part of the image you want to change. Want to adjust, give it some. New instructions, and skip everything else.
This overcomes a major limitation of AI art, which is that every photo (and video) is completely unique and different, even when you’re using the same gestures. This makes it difficult to achieve consistency across images, or to nail down an idea. However, based on these AI art creators, known as Diffusion modelsThere are still many limitations to overcome—as we’ll show you here.
Editing images in ChatGPT
If you’re a ChatGPT Plus subscriber, you can load the app on the web or mobile and request a picture of anything you like: a case-solving cartoon dog detective in a cyberpunk setting, Lonely Hills. A rolling scene. Image of middle distance and gathering above storm clouds, or whatever. After a few seconds, you will get your picture.
To edit the image, you can now click on the generated image, and then on select button in the upper right corner (looks like a pen writing a line). You then adjust the size of your selection tool using the slider in the top left corner, and drag over the part of the image you want to change.
ChatGPT editing interface
Credit: Lifehacker
This is where this is an important step: you can leave part of the image untouched, and only refresh a selection. Previously, if you sent a follow-up prompt to change a particular part of the image, the entire image would be recreated, and probably look quite different from the original.
When you’ve made your selection, you’ll be prompted to enter your new instructions, just for the featured image. As always with these AI art tools, the more specific you can be, the better: you can ask a person to look happier (or less happy) or paint a building a different color. Your requested changes are then applied.
Success! ChatGPT and DALL-E replace one dog for another.
Credit: Lifehacker / DALL-E
Based on my experiences, ChatGPT and DALL-E seem to do the same kind of AI tricks we’ve seen with apps like Google’s Magic Eraser: intelligently filling in the background based on information in a scene. , while trying to leave everything out, the choice untouched.
It’s not the most advanced selection tool, and I noticed inconsistencies in borders and object edges—which is to be expected, considering how much you have to do when it comes to making selections. Control is achieved. Most of the time the editing feature worked fairly well, although it’s by no means reliable every time, which OpenAI will no doubt try to improve in the future.
Where AI Art Hits Its Limits
I tried out the new editing tool to do various tricks. He did well in changing the color and position of a dog in a meadow, but less so in reducing the size of a giant man standing on a castle rampart—the man was only a fragment of the rampart. Disappeared into a blur, which suggests that the AI was trying to paint around it without much success.
In the cyberpunk setting I asked the car to go in, and there was no car in sight. In another castle scene, I requested that a flying dragon be rotated so that it faced the other way, turned from green to red, and had flames coming out of its mouth. After a few moments of processing, ChatGPT removed the dragon completely.
failed! ChatGPT and DALL-E erased the dragon instead of replacing it.
Credit: Lifehacker / DALL-E
This feature is still fairly new, and OpenAI isn’t claiming it can replace human image editing—because it clearly can’t. It will improve, but these mistakes help show where the challenges lie with certain types of AI-generated art.
What DALL-E and similar models are very good at is figuring out how to arrange pixels to approximate a castle (for example), based on the millions (?) of castles they have. Training has been given. However, the AI doesn’t know what a castle is: it doesn’t understand geometry, or physical space, which is why my castles have turrets sticking out of nowhere. You’ll see this in a lot of AI-generated art involving buildings, or furniture, or any objects that aren’t perfectly rendered.
It’s very white but it’s far from “plain”.
Credit: Lifehacker / DALL-E
At their core, these models are probabilistic machines that don’t understand (yet) what they’re actually showing: that’s why, in many of OpenAI Sora’s videos, people never disappear. , because AI is very cleverly arranging the pixels, not tracking people you may have also read about AI. Struggling to make pictures of pairs of different races, because based on the image training data, there are more pairs of the same race.
Another quirk that has been noticed recently is the inefficiency of these AI art generators Create a plain white background. These are incredibly smart tools in many ways, but they don’t “think” the way you or I would, and they don’t understand what they’re doing the way a human artist would— And it must be endured. Keep in mind when you use them.
Credit : lifehacker.com