If you’re familiar with OpenAI, one of the biggest artificial intelligence companies in the world today, you’ve probably heard of DALL-E 3. Maybe you’ve even experimented with previous versions of OpenAI’s image generator application – but learning how to use DALL-E 3 is slightly different.
Like most of OpenAI’s products, including the world-changing ChatGPT, DALL-E has gone through numerous iterations and evolutions over the years. Today, it’s easier to access than ever before (provided you’re willing to splash out on an OpenAI subscription).
Here, I’ll walk you through everything you need to know about DALL-E 3, from how it works and what it can do to how you can start using the app to create unique visuals.
What is DALL-E 3?
DALL-E 3 is the image-focused generative AI model designed by OpenAI. It’s similar to ChatGPT in that it “creates” content based on user input (such as text-based prompts). The main difference is that DALL-E is entirely image-focused.
DALL-E 3 can’t create blog posts and social media captions for you like ChatGPT, but it can produce high-quality, unique images for your social media accounts, blog posts, and more.
Fun fact – the creators came up with the name by blending the famous surrealist artist (Salvador Dali) name with the Pixar movie character WALL-E.
OpenAI first launched the initial version of DALL-E in January 2021. At the time – it was revolutionary – one of the first generative AI applications to show us just how creative and artistic large language models could be. The first version of DALL-E used technology called a Discrete Variational Auto-Encoder, based on research from Google Deepmind.
One year later, OpenAI built on the existing model with a successor: DALL-E 2. This updated model leveraged more advanced techniques to boost realism, such as a stable diffusion model integrated with Contrastive Language-Image Pre-Training (CLIP) model data. The CLIP model was trained on over 400 million labeled images, making it more effective at creating valid responses to user prompts.
Finally, in September 2023, OpenAI announced DALL-E 3. This version of DALL-E is now available via ChatGPT, and in some Microsoft products.
What’s New with DALL-E 3? What Can it Do?
According to the team, DALL-E 3 is a significant upgrade to the model, similar to the difference between GPT 4, and GPT-4o. The new model is more effective at understanding significant nuance and detail than its predecessors and follows complex prompts more accurately.
Compared to the older models, DALL-E 3 benefits from:
- Enhanced context understanding: Compared to the predecessor models, DALL-E 3 understands more information and context. It’s much more effective at producing visuals that adhere to textual descriptions provided by the user. It’s also less likely to create highly unusual or bizarre images by accident (though this can still happen).
- Direct Integration with ChatGPT: Unlike previous models, which were available as standalone apps, DALL-E 3 is integrated into ChatGPT—at least for the time being. This means you can benefit from communicating with ChatGPT naturally while still having the option to rapidly create images from conversations.
- Safety protocols: OpenAI, like many AI innovators, has increased its focus on developing safe and ethical AI in recent years. DALL-E 3 prohibits the generation of aggressive or discriminatory images. It also allows artists to determine whether they want to exclude their images from the materials that the app can use for reference.
How to Use DALL-E 3 with ChatGPT: Getting Started
There are currently a couple of ways you can use DALL-E 3. Thanks to OpenAI’s continued partnership with Microsoft, DALL-E 3 is available in certain Microsoft apps. For instance, in Microsoft Designer, Copilot can use DALL-E 3 to create images for users in the “flow of work.”
You can test DALL-E 3 out in the Image Creator tool within Designer, but you won’t have the same level of control over the system, and you’ll have to keep track of your “boosts” credit system.
If you’re not a Microsoft customer, however, the easiest way to use DALL-E 3 is through ChatGPT. The only caveat is that, for now, DALL-E 3 is only available to subscribers using either ChatGPT Plus or ChatGPT Enterprise. Alternatively, you can access it through OpenAI APIs if you’re a developer, but there’s no option for this model on the free version of ChatGPT (yet).
To use the app, you must sign up for a ChatGPT account, then click the “Upgrade” button to switch to a ChatGPT Plus or Enterprise account.
For now, the image generator seems to have limitations similar to GPT-4o on paid accounts. On the “Plus” account, you can send only 80 messages every 3 hours to GPT-4o. For DALL-E 3, you may be limited to up to 50 requests every 3 hours. However, this could change in the months ahead. Plus, it’s worth noting that it’s more than you’d get with most image generators.
Options like Midjourney and Dream Studio powered by Stable Diffusion have much lower limits.
How to Use DALL-E 3: Creating Content
Once you have your premium ChatGPT account, log in on your chosen browser, and make sure you’re using ChatGPT 4 or ChatGPT 4-o. Alternatively, you can visit the DALL-E 3 GPT model here. From that point, using DALL-E 3 is a lot like using ChatGPT.
All you need to do is enter a prompt into the text bar to start creating content. You can type in your text or use natural language (your voice). If you’re using the link above, you’ll get two images for each prompt you enter. Using the standard ChatGPT browser application, you’ll only get a single image for each prompt.
There’s no limit to what you can enter, but it’s a good idea to be specific with your prompt, outlining precisely what you want to see and the style you want for your art. For instance, try something like “An oil painting of a dog drinking a pina colada on the moon.”
If you use the dedicated GPT, the interface also includes some suggested prompts and an aspect ratio option.
Notably, one thing that this model does better than its predecessors is rapidly improve and iterate on the prompts you enter. Instead of just responding to one prompt in the same way for both images it creates, it looks at the prompt from different angles. This means you get more variation to choose from. Plus, DALL-E 3 allows you to check out the sources for your content.
If you click on a generated image and then the “I” button, it will show you what the bot used to respond to your prompt—which is handy if you’re trying to avoid plagiarism issues.
Downloading and Editing your DALL-E 3 Images
If you’re happy with the image the app creates, you can save it to your computer by clicking the download icon at the top right corner. However, most people learning how to use DALL-E 3 for the first time will need to consider editing.
We’ve all seen examples of weird and wacky images created with AI that don’t make a lot of sense. This is partially to do with how AI works – the connections AI systems make between words and prompts aren’t always as linear as we might expect.
DALL-E 3 can occasionally create unusual content, but it gives you more scope to edit whatever you create. If you want to change something in the resulting image, all you need to do is tell the app what you want to adjust. For instance, you can:
- Ask for a few different variations of an image
- Alter the point of view for your image
- Change where the subject appears in your image
- Add and remove background details
- Modify details like colors and sizing
- Change the aspect ratio of the image
- Add or remove subjects from an image
Controlling Changes with DALL-E 3
Keep in mind that when you direct the bot through this process, it builds a new prompt based on your input and generates new images rather than making direct edits. Sometimes, the changes you see will be minimal, while other times, they can alter your output entirely.
Although the app usually doesn’t make drastic changes, it could end up removing the parts of your image that you liked the most.
If you want more control over the editing process, click the image and Select tool in the top menu. You can use this to paint over the area of the image you want to edit. You’ll also be able to adjust the brush size with the menu. In the sidebar, prompt DALL-E 3 with what you want to change in the specific section you highlighted.
How to Use DALL-E 3: Top Tips for Better Results
Learning how to use DALL-E 3 is simple enough, particularly now that it’s included in ChatGPT (for paying users) and certain Microsoft apps. However, it can be tricky to ensure that you get the best results from your prompts every time.
Like most current LLM solutions, the technology for this app is still being developed, and the algorithms are still learning. That doesn’t mean you can’t get great results, however—we think the paintings and drawings created by the bot are pretty impressive.
Still, if you’re looking for ways to boost your chances of success, here are our top tips.
Practice Detailed Prompting
Prompt engineering is an art and a science that many people are still getting used to in the world of generative AI. You don’t necessarily need to be a technical expert to create great prompts; all you need is the right level of detail.
Remember, you’re giving this bot instructions to create a highly detailed image. Many different factors can affect the outcome, from the artistic style used to the coloring, the subjects, and the aspect ratio. Be specific with what you want.
Instead of just typing “Give me a surrealist picture of a cat”, try something like: “Create an image of a ginger cat in the surrealist style, sitting on top of a table, with a meadow behind it. Make the day appear sunny and create a sense of joy in the picture.”
Remember Numbers and Positioning
Although providing an app like DALL-E 3 with too much information is possible, it’s much harder to overwhelm the algorithm than it was with the previous model. Although this application isn’t perfect, it’s much better at understanding specific details, like numbers and the positioning of elements.
For instance, you can ask it to generate something in the left or right-hand side of your image, and ask it to make the subject 30% larger than an object in the background (like a tree).
Ask for Subtle Edits
If you ask the bot to edit the images it produces for you, you won’t always want massive changes. If you’d prefer to keep things similar, but just change a few specific elements, make sure the bot is aware of this. Ask for “subtle variations” or tell it to keep certain features the same.
Remember, you can send numerous requests to the algorithm within three hours, so don’t be afraid to make small changes at first. Creating your images iteratively, through conversations with the bot, should increase your chances of getting the results you want.
Start Experimenting with DALL-E 3
DALL-E 3 is definitely a major upgrade over the previous image generation apps offered by OpenAI. It’s much better at understanding complex prompts and creating highly realistic images. However, it still suffers from occasional bugs.
For the most part, the best way to get a sense of what this tool can do is to experiment with it yourself. If you have a premium ChatGPT account or use Microsoft Designer, you can use the model today to create custom images.
If OpenAI continues to build on its DALL-E collection with more iterations, we could even see the third version of this model becoming available to free users in the years ahead.
