Unsure of how to use DALL·E 2? Perhaps the best-known of the latest AI art generators DALL·E 2 can allow anyone to generate images of practically anything merely by writing a text description. At least that's the theory, but the platform has a few quirks.
DALL·E 2 was first launched in April 2022, and it immediately caused astonishment due to the apparent accuracy and realism of the images it can create from text prompts. Access was initially via a waiting list, but the tool, although still in beta, is now available for the public in general.
As with all AI art generators, the tool remains controversial due to fears of potential misuse, the possible impact on creative jobs and disregard to the copyright on the images used to train the model. Those concerns aside, many artists and designers are busy exploring how to make this new technology work for them.
How do you use DALL·E 2? Well, in theory, you go to the website, create an account and you can generate images of anything you can image by typing a text prompt. However, there are a few things to bear in mind. Below, we'll walk through the process of how to access DALL·E 2 and how to use DALL·E 2, including some tips on how to make DALL·E 2 images better (if you prefer to create your own original work in a more traditional way, see our guide to the best graphic design software).
How to use DALL·E 2
DALL·E 2 is a browser-based tool (it also offers an API that developers can use in their own apps). To use DALL-E 2 directly, first off, you'll need to create an account. This is a quick process and once you've confirmed your email or telephone number, you use the tool straight away. Everyone gets 50 free credits to start with, and that's enough to explore how the tool works. You get another 15 free credits each month, but they don't go very far, so if you want to use the tool regularly, you'll eventually need to buy more (see below for more on the pricing).
Once you've completed the DALL-E 2 sign up to open an account, you use the program in your browser on the DALL-E 2 website (opens in new tab). It's fairly intuitive at first. Type in a description of what you want in the text box, for example 'an astronaut riding a horse in an impressionist style'. Click 'generate' and DALL·E will attempt to create four 1024x1024 images representing what you wrote.
You may have to tweak your prompt to get a result you like. When you have an image you're happy with, click on the one you like. On the next screen you can download the image (click the arrow in the top right of the image). You can also edit the image (click 'edit'; this gives you tools such as an eraser and options to upload an image to add to the AI creation or to add more squares to expand your image). You can also create 'variations' of the image – DALL-E 2 will create 4 alternative version of the image you created.
How to write prompts for DALL-E 2
It sounds simple, and it is. But getting good results can be tricky. Prompts that a short on detail tend to be haphazard, so it makes sense to try to make the most of the 400 characters you can use. It often seems like the more detail the better. For example, if you're aiming to create photorealistic images, it can help to include the model of a camera, the focal length of a lens and the type of lighting (think 'studio lighting', 'soft lighting', 'deep shadows', etc).
Until you get used to how much detail you need to include in the description, you'll often find that you need to edit the prompt and try again after the first results. In the first set of images of the marmot skiing above, the marmot looks kind of terrified. Perhaps, I want him to look like he's enjoying himself, so I'd need to change that by editing the prompt. I could add 'smiling'. Now he looks happier... kind of... in some of the options. Be careful, because each new generation will cost you a credit.
How to use DALL·E 2's other features
The other way to use DALL-E 2 is to upload an image from your phone or computer. You'll find a link to click to upload an image right below the prompt field. The image will be cropped to a square. Once you've uploaded the image you want to use, you can have DALL-E 2 create its own variations of the image or you can edit the image. In the example below, I uploaded an image of a girl cycling past someone carrying an umbrella. With no further input DALL-E 2 generated some interesting variations, including some that combined the two subjects to show someone riding a bike with an umbrella.
How to use DALL·E 2 outpainting
DALL-E 2 outpainting lets you expand an image beyond its original borders. This can be used with an image created in DALL-E 2 itself or an image you've uploaded – people have been using it to expand famous works of art, such as the Mona Lisa, adding more background to the image.
To do this, in the editor generate or upload an image as above, and then drag the corner of the image to reduce its size in the generation frame. Write a prompt (this time it can be more general because DALL-E 2 will try to match the style of the existing image – in the example below, I wrote 'mountainous landscape'). DALL-E 2 will then fill the rest of the frame with something that will – hopefully – fit your image.
How to improve DALL·E 2 images
It's unlikely that DALL·E 2 is going to give you the results you want the first time. You might have to edit images by writing new prompts for specific elements within them. And there are somethings that may just be impossible to correct in DALL·E 2.
There may be unwanted artefacts in images, and, like all AI art generators, DALL·E 2 can struggle with human faces and bodies. To improve images created in DALL·E 2 you may need to export them and then edit them in more traditional image editing software like Photoshop. For example, you can fix wonky eyes, using Photoshop's Neural Filters.
DALL·E 2 also has a limit on resolution, which means it isn't possible to create images that are big enough to print at large sizes, but you may have success upscaling your image in another program. See below for more common questions about DALL·E 2.
How can I make my DALL·E 2 images better?
If your DALL-E 2 images aren't turning out how you'd like there are a few things you could bear in mind. Ensure your prompt is a detailed as possible – including detail about the background if that's important. Specify the style of art you want if you're going for an artistic style, or specify camera angles, lighting details, focal distance and even the specific camera if you're aiming for a photographic style.
DALL-E 2 struggles with requests for images of multiple subjects, so if possible keep it to one subject. That can be limiting, but if you want to create images with multiple subjects, you could try creating separate images and then editing them together. Remember that you can edit images in Dall-E 2 or alternatively export them to edit in the best photo editing software.
DALL·E 2 prompt book
Guy Parsons has created a DALL·E 2 prompt book (opens in new tab) as a visual resource to help inspire prompts in DALL·E 2. Published by the AI art website the DALL·Ery GALL·Ery (opens in new tab), it offers some useful pointers on how to get the most out of the AI art generator.
The 82-slide guide covers techniques for eliciting the results you want from DALL·E 2 It covers aesthetics and vibes, recommended adjectives to use to get the feel or composition you're after and tips on everything from photography to painterly portraits and landscapes, film and illustration styles and 3D art. For illustration, it provides examples of different styles, media and textures. It also shows how you can use styles from art history in prompts to get interesting results.
The guide points out that even the creators of DALL·E 2 don't know what the tool knows and doesn't know. Instead, users have to work out what it's capable of doing and how to get it to do what they want. One piece of advice is to be specific – state whether you want a close-up image or a particular angle. And remember that an adjective without further definition could be interpreted in different ways – to influence the whole look of an image or something as specific as the style of dress of the subject.
What is DALL·E 2?
DALL·E 2 is a text-to-image AI art generator based on machine learning that's been provoking both horror and awe online (see our pick of the weirdest AI art created by DALL·E 2). Created by the artificial intelligence company Open AI, it's a generative tool, which means it can generate art from scratch as well as create edits or variations of existing work. It doesn't actually 'know' what it's creating but it makes assumptions based on the massive database of 650 million image and caption combinations that it's already been fed.
The name is a portmanteau of 'Dali' (as in Salvador) and Pixar's 'WALL-E'. As the name suggests, this is the second iteration of the tool, and it seems to be a major improvement on the first, which tended to produce grainy images, and take a long time too.
It's not only generative AI art creator to work based on text prompts. Artbreeder has launched Artbreeder-collages, which blends text prompts with a collage-like design process. Stable Diffusion and Midjourney are also popular. What appears to potentially set DALL·E 2 apart is the results for particular styles of image, particularly more photorealistic images (see how the best AI art generators compare).
Is DALL·E 2 available to the public?
For the first five months after the tool's release in April, DALL-E 2 access was limited, and there was a long waiting list. But in September 2022 access has been opened so that anyone can complete DALL-E 2 sign up (opens in new tab).
According to OpenAI, “Responsibly scaling a system as powerful and complex as DALL-E – while learning about all the creative ways it can be used and misused – has required an iterative deployment approach.” It says it's now shored up its safeguards enough to open access to everyone.
The catch? DALL·E 2 is no longer free to use. Instead, users will be given a limited number of monthly credits, with the option to pay to top them up (see below).
How do I sign up for DALL-E 2
You can complete DALL-E 2 sign up by creating an account on the Open AI website. You'll be asked to enter your email address and a security code and to create an eight-digit password. You'll then receive an email with a link that you'll have to click through to verify and you'll also receive a text code by SMS that you'll have to enter to confirm your identity. You can also create an account using the likes of Google or Microsoft. Click 'continue' to accept the terms and conditions, and you're ready to get started using DALL-E 2.
How much does it cost to use DALL·E 2?
Is DALL-E 2 free? Not any more. It was free until July 2022 (for those that had access), but OpenAI now uses a credit-based model. New DALL·E 2 users get 50 free credits that they can put towards generating, editing or creating a variation of an image (new image generations return four 1024 X 1024-pixel images for the cost of one credit).
After that, users get 15 free DALL-E 2 credits each month. To get more, you have to buy them at a price of $15 for 115 credits (enough to generate 460 1024 X 1024-pixel images). OpenAI has invited artists who need financial assistance to apply (opens in new tab) for subsidised access.
Is there a free DALL·E 2 alternative? Yes, there are several free AI art generators available. As well as Art Breeder-Collages, which we mentioned above (currently in beta), there's DALL-E mini – now called Craiyon (although it produces some nightmarish results), and Stable Diffusion, which is open source.
Can I use DALL·E 2 for commercial use?
Until now OpenAI had prohibited commercial use of images generated by DALL·E 2, but in the beta version, it's now giving “full usage rights” for images created with the platform. That includes the right to sell and reprint images and to use them on merchandise. We have already seen the first case of someone copyrighting AI work.
However, there are concerns about the copyright implications of training an AI model on existing images. Getty Images has banned the use of AI content on its stock image library out of concerns that copyright is murky, and some politicians are calling for specific legislation to clear the matter up.
Are there restrictions on DALL·E 2?
OpenAI says that it's able to start expanding access now thanks to changes in its policies and advances in mitigating "bias and toxicity" in images generated by the platform. The company says that this week it made a change that will push DALL·E 2 to generate images of people that “more accurately reflect the diversity of the world’s population” if race or gender is not specified in the text prompt.
It says that it's also taken steps to ensure that the platform rejects image uploads that contain realistic-looking human faces or the likeness of public figures, such as politicians and celebrities. OpenAI says it doesn’t allow DALL·E 2 to be used to create images that could cause harm, for example, images showing self-harm, hateful symbols or illegal acts. It stressed that it has both automated and human monitoring systems to prevent this, as well as to prevent DALL·E 2 from memorising faces that appear a lot online, however it recognises that there's more work to be done in this area.
How can you tell if an image was created by DALL·E 2 AI?
You can tell is an image has been created by DALL·E 2 because they contain a signature that looks like a row of coloured squares at the bottom right of the image (assuming the image hasn't been cropped. See the example above.
Can you remove the DALL-E 2 watermark?
When you download an image created in DALL-E 2 it will have the colour swath watermark in the bottom right of the image, however according to DALL-E 2 terms this can be removed – which in many cases might be necessary for commercial work.
It should be fairly easy to remove the watermark in any app a with an object removal, clone stamp or content-aware fill tool, for example Photoshop. There is also a way to directly download the image without the watermark. On desktop, you can right click the image, choose 'Inspect' and then look for the windows.net URL. Copy this image link and you should find it doesn't contain the watermark. On mobile you can tap and hold the image on the generation page and click 'save image'.
Why is DALL-E 2 controversial?
There are concerns that AI image generators could be misused because people may pass off images created using them as real, leading to fake news. There are concerns that they could also be used to create violent images or non-consensual porn. OpenAI says that it's taken precautions to avoid this and continues to learn from real-world use while the program remains in beta.
Some people are also concerned about what AI image generators mean for the future of creative jobs. Our take, for now, is that shouldn't be putting any artists out of work, but rather that there may be potential for creatives to make the tool work for them.
How do I use the DALL-E 2 API?
Developers can integrate DALL·E directly into their apps and products using the API. For example, Microsoft an investor in OpenAI, is using DALL·E 2 for a new graphic design app called Designer, which aims to cover social media design, invitations, digital postcards and more. Microsoft is also integrating DALL·E 2 into Bing and Microsoft Edge with Image Creator, which will allow users to create images if search results don’t come up with what they’re looking for. You can learn more in the documentation (opens in new tab) on the OpenAI website.
Read more:
