How to Use the Dall-E AI Art Generator to Create Stunning Images from Text

AI art generators have been in the news a lot this year, whether for their amazing advances or questionable uses. OpenAI’s Dall-E 2 is one of the leading names in this space. It is now open to the public and developers, and will soon be integrated with Microsoft software and the Bing search engine.

Stock art service Shutterstock will also integrate the tool and pay the artists it copies to give back and (perhaps) avoid some of the ethical concerns. After all, Shutterstock’s art was used in part to train the Dall-E AI.

But how, exactly, do you work with Dall-E? Is it really as simple as typing in a description (called a prompt) and fetching an image? To be honest, yes. But there’s a lot more to it if you want to get closer to the perfect result. Let’s talk about it.

Register and pay

The first thing to do is to create an account on labs.openai.com(Opens in a new window). Use a Google or Microsoft account, or create a login with an email address and a strong password. There is no multi-factor authentication option.

Dall-E is not entirely free. The service operates on “credits(Opens in a new window). » You get 50 free credits when you sign up, then 15 free credits per month thereafter, but they don’t roll over. Paid credits are rolled over from month to month, up to 12 months; get 115 credits for $15.

One credit lets you make one AI art build (you get four new images per typical build). It could start from a prompt, but a credit is also used to create a variation of already generated art. You can use a lot of credits trying to crawl the right AI-generated image.

The perfect invite

Once registered, you will be faced with a form for your invite. If you click on the Surprise me you’ll see other random prompts drop into the text box – they won’t count towards your credits until you click Generate. You can also upload your own image and use Dall-E to modify it to add new AI-generated content or to create all-new variations of the original.

(Credit: PCMag)

But the prompt is the thing. This is probably where most people will stumble. We did our own experiments to see how random prompts can do with Dall-E and competitors like Midjourney(Opens in a new window) to create art. The initial prompt always does something… interesting. But it’s rarely perfect. Things are still a little soft. Bit strange. A little off. The art turns out to be much better if you perfect the prompt, which is limited to 400 characters (emojis(Opens in a new window) work(Opens in a new window)).

Design prompts(Opens in a new window) have both contents (what you want to see) and modifiers (how it should look like). For example: “A robot drawing a painting on an easel” is content, but “Over the shoulder view, colorful, oil painting, in the style of Van Gogh” are all modifiers.

Not feeling very creative? There are people out there who will help you create the right prompts to get the right art. They carry job titles such as Prompt Designers or Prompt Engineers, and their expertise does not come for free. But if you’re going to do this often, getting the prompt right off the bat can save you some money, since you won’t be using as many credits.

In addition to videos like the one above, there are sources like The Dall-E 2 Prompt Book(Opens in a new window) from dall-ery gall-ery. It is worth reading. It spells out many “hacks” you need to consider in your prompt, such as using terms like “close-up” and possible camera angles, types of lighting, list of eras to mimic, such as “1920s”, or even mention of a specific camera lens or type of smartphone that “takes” the AI ​​image.

Using emotional words to get a more positive or negative image is important, as is using words to set the mood/aesthetic(Opens in a new window). The prompts for getting the most out of AI-generated “photos” and “artwork” is a whole world unto itself. The possibilities are limitless. Your only limit is 400 characters.

The only thing you can’t do with Dall-E is generate images of real people. Well, you’re not supposed to, anyway. It has filters and limitations in place to make sure of that, but with billions of images studied, who knows when Dall-E might post an image that looks like a celebrity. That said, you can reference movies and shows to get closer to their look and feel.

Can you ask for a style that mimics a living artist? Sure, and if Dall-E has seen this artist’s work, he can probably deliver a knockoff. A style can’t be copyrighted, but of course that’s a shitty thing to do if you’re going to use it commercially when you could have been working with an artist.

Image variants

Any image you generate in Dall-E – or any image you upload to Dall-E (make sure you are the copyright holder) – can get instant variation. Uploaded images should be cropped into a square image at 1:1 ratio.

(Credit: PCMag)

You can also create variants of a variant. Just keep clicking the three-dot menu on the images and select Variation to do more.

Variations cannot really be controlled. It’s just Dall-E looking at the content of an image and creating something approximately the same, with the same styles, similar layouts, etc.

Let’s say you created an image with Dall-E and you like it. Most. But there is something wrong. Select Edit, and remove the part you don’t like with the eraser tool, rewrite part of the prompt to address that section (eg a face that needs a more specific description or a whole new background to replace the ‘original). Dall-E creates a variant based on the new prompt specifications. Your prompt should always describe the entire desired image, not just the erased area.

This eraser option also works with images you can upload, so you can remove a head or an entire background. Write a prompt for what you want to appear in the erased section. For example, erase some sky and use the “upside down plane” prompt to see it appear overhead. Working within the boundaries of an image generated in Dall-E is called “inpainting”.

Delete part of your download, create a prompt and generate new AI art there. (Credit: PCMag)

Enlarge an image

So what is “overpainting”? Well, usually the output of Dall-E is limited to uploading an image measuring 1024 x 1024 pixels. Another thing you can do under Edit is to create generation frameworks. Click the Add Build Frame icon that looks like a box with a plus sign in the top left (or press the letter F), and you’ll get a floating box that you can place anywhere outside the perimeter of the image.

Generating a frame to extend a Dall-E image. (Credit: PCMag)

Click Generate again and the image will expand into that frame, as if the AI ​​just continued to draw/draw into the expanded space with the same prompt. Continue until the image is as large as you want. Each new image generated uses a credit, of course.

When you’re ready to own it, click the download button (the down arrow) and you’ll get a .PNG file of your new masterpiece. If you really want to increase the resolution of your image after that, you can try a website like bigjpg.com(Opens in a new window) or ARC Face Restorer(Opens in a new window) (which also fixes faces) or commercial software like the $99.99 Gigapixel AI(Opens in a new window).

Don’t forget that you can always go back to your My Collection page on the Dall-E site and see all of the images you’ve created, including variations. So if you have an idea to improve an old image that you never uploaded, you can come back to it.

If you’ve created something really amazing, you can click Share > Publish. This places the image on a public page on openai.com, for which you get a special link(Opens in a new window) to share with others. It will display the image along with the prompt used to create it. If you don’t want to be so public, try Save > Favorites to put it in a collection that you can see later.

Become an artist

Ready to try your hand at the art of AI? All you have to do is sign up and type. Try it out at Dall-E with the free credits and you’ll get a taste of all the AI-generated art that’s likely to come your way in the coming decades.

A variation on a download, extended with new frames. (Generated on Dall-E)