6 Tips To Create Images With DALL-E 2 + Prompt Examples
Technology continues to prove that creativity truly knows no bounds. If you have used DALL-E 2 to generate images, you know what we are talking about. A panda on an ice rink, a futuristic landscape with skyscrapers painted in Art Nouveau style – you can let your imagination run wild and DALL-E 2 creates them for you, in seconds. The finesse of tools like these has made people rethink the skepticism surrounding AI and has made people more accepting of such emerging technologies.
Can you believe that the John Audubon painting below was generated by AI? The below DALL-E 2 generated image is proof that AI tools are causing quite a stir in the world of creativity.
Though DALL-E 2 is pretty easy to use, to get images that look like this, you need the best prompts. So, in this blog, we’ll talk about the tool itself and some useful tips to help create stunning images with it.
DALL-E 2 an overview
OpenAI introduced the world to DALL-E back in January 2021. This was the time when you must have caught a lot of people talking about an avocado armchair!
By 2022, DALL-E 2 was out. The biggest leap was with the four times larger image resolution support. It also made heads turn with its stunning photorealistic results.
Initially, DALL-E 2 was only available to limited users. But after months of phased deployment, OpenAI finally put an end to waitlists and made DALL-E 2 available to everyone by September 2022.
Image generation is undoubtedly the signature feature of DALL-E 2. But there’s a whole lot more you can do with it. Let’s look at some features to explore.
Variations
For every prompt, there are four images generated. If you like one of them, you can then create variations of one of these images from the Quick Actions menu displayed on the respective image.
This option generates four more variations of the selected image which you can then download or generate more variations. Here’s an example.
These variations come in handy especially to create simple animated GIFs.
Inpainting
Another useful feature is Inpainting. This is where you edit an existing design and retouch it based on your requirements. This feature helps you add elements or remove them from the image without distorting the background and affecting the other elements present. This is a great tool to add textures or shadows to an image or even remove them.
Outpainting
The below video shows how DALL-E 2 outpaints (expands) a version of the popular artwork, Girl with a Pearl Earring by Johannes Vermeer.
This feature is great for reimagining existing art to understand how AI interprets art and processes it. And it helps you creatively tweak the resolution of your existing design for use on a different platform or for use in print.
Getting started with DALL-E 2
One of the best things about DALL-E 2 is its simplicity. It is available as a web application that you can directly access from your internet browser or use as an API in case you wish to integrate the AI image generation function into your app.
Start by creating your free account. Once you sign up you will see 50 credits added to your account and there will be an additional set of 15 credits added on a monthly basis. This is for the free account. And the free credits cannot be carried forward.
You can also choose to purchase credits. Remember that the pricing and credit usage vary for different models on OpenAI.
As you can see from the image above, the price and therefore the use of credits in your account depend on the resolution of your image. Paid credits are available in your account for 12 months before they expire.
Once you create your account, the process is pretty simple. You key in your prompt and click Generate or upload an image to start editing.
DALL-E 2 prompt – 6 tips to get the best results
With AI image generators like DALL-E 2, you either get stunning results that blow your mind or funny-looking renditions of what you had in mind. The difference lies in the prompts you use. Your prompt can be up to 400 characters long. Make the most of this character limit and add in as many details as you can in order to get the best results. Let’s look at some tips to create perfect DALL-E 2 prompts for better image generation.
1. Get specific about the style
You need to tell AI what kind of visual style you have in mind. Otherwise, you get a generic rendition of the keywords.
“A cityscape” is simple enough for the algorithm to understand what you are looking for. But what about the visual style and mood of the design? Clearly specify your preferred style in the prompt.
Do you want a watercolor illustration or perhaps an oil painting style? Or do you want your image to resonate with a popular art movement like Impressionism or perhaps mimic the style of a popular artist? Get more specific about the style and the design is better.
The below images were generated on DALL-E 2 for “cityscape”. Notice the difference adding a description of the visual style can make.
Kimp Tip: Within every design style there are subcategories. For example, illustrations can be of various styles. It can be an analog style like a pencil drawing or watercolor illustration, digital styles like vector art, or low poly. And when you have to talk about art styles, there are options like Afrofuturism, steampunk, Memphis style, and so on. When you know what each style represents, you know the exact terms to include in your prompts.
For example, both the below images were generated for “painting of a flower vase on a kitchen table” but one had an additional detail – Memphis style.
2. Talk about every little element in your image
One common mistake that people do while using AI image generators like these is that they talk about the subject but forget about the background.
In the above example, a simple prompt like “a teddy bear on a skateboard” would have resulted in an image with a plain white background. Adding the location detail adds to the accuracy of the image. So, when you create your prompts ensure that there are enough details about the main subject of the design as well as the background or the scene in which you would like your subject to be present.
3. Don’t forget to talk about colors
Color of the subject’s costume, background colors, or an overview of the color moods you would like can add value to your prompt.
In the above example, had the background and subject colors not been specified, the image would have looked so much different.
To further enhance your results, you can add the right descriptors to your color names like “bright pink” or “sky blue” or even use specific color names like “crimson red” and “magenta”. Better yet, if you have the HEX code of these colors, you can add them to your prompt to get the exact colors you have in mind.
4. Talk about emotions
Do you want your subject to be happy, sad, angry, or perhaps lost in thoughts? Describe the emotions you wish to incorporate. After all, without a strong emotional quotient, your image lacks depth.
The difference between the below images generated by DALL-E 2 is just the inclusion of the word “happy” in the prompt.
Adding emojis can also help capture the expression more accurately when you have animals or human beings in the picture or even anthropomorphic subjects.
You can even create prompts with just emojis. But yes, the interpretation varies drastically and you might not always get the results you imagine.
5. Describe the composition of the image
With human subjects and animals, specifying emotions and expressions is easy. But what if there are only inanimate objects in the image? These images can still have emotions. After all, that’s how you connect with your audience effectively.
For this, you can add details about the composition of the image, parameters like:
- Photo angles – a macro shot of the subject, or an overhead shot or a drone shot.
- Lighting – studio lighting, indoor lighting, outdoor lighting, backlit shots.
- Photo lens effects – fisheye lens, double exposure
All those little tweaks you might be familiar with, all the parameters that you usually adjust in a photo editing app can make valuable inputs in your DALL-E 2 prompts. These help craft the scene accurately and deliver a composition that matches your idea.
6. Avoid complicated prompts
Descriptive prompts deliver great results, no doubt. But adding complicated details and very accurate dictation of elements like orientation and position of objects in the design might not always simplify things. You might not get the results expected with prompts that are too specific.
For example, we tried creating a design with a prompt that talks about the position and layout of the elements. “On the left side, a yellow rectangle slanting toward the center and on the right a big red circle placed at the top right corner”. Though the prompt reads simple the results were not quite accurate.
Similarly, DALL-E 2, AI image generators in general, might not be the most suitable option for creating images with multiple human subjects or groups of people. In several designs, the face appears distorted and by experimenting with tweaking the prompt for editing the features you are spending your credits in vain.
With these do’s and don’ts in mind, go explore the potential of DALL-E 2 and find inspiration for your marketing designs or create unique elements to add to your ads. That’s how you put AI to good use in complementing the efforts of your creative team.
How does DALL-E 2 compare with Midjourney?
Midjourney is another popular AI image generation tool, a close contender to DALL-E 2. If you are looking for details about the pricing and prompts for Midjourney, check out our blog here.
To quickly sum up the differences between these two popular AI imager generators:
- Midjourney is currently only available through a Discord server whereas DALL-E 2 can be used directly on your browser and in the form of API.
- Midjourney licensing terms for images depend on the plan you choose. And there are monthly charges billed depending on the plan. With DALL-E 2, instead of the subscription model, you get a pay-as-you-go model where you buy credits depending on your usage. And images created with free credits as well as purchased credits can be used for commercial purposes.
In terms of the designs generated, both Midjourney and DALL-E 2 deliver some pretty impressive results. Both tools give you four different variations to choose from. Outpainting option on DALL-E 2 feels much more practical than the upscaling option on Midjourney. Moreover, the former also gives you the option to edit your images before you download them.
Summarising all these differences and similarities, DALL-E 2 feels like a slightly more convenient and feature-packed tool for beginners exploring AI image generation.
Redefining your creative process with a combination of AI and the right design outsourcing strategy
Brands and content creators around the world have been using DALL-E 2 and other such AI image generators to create creative and cool designs that can be used in digital stickers, AR technology, games, and more.
How do you use AI-generated designs? Well, remember that these tools only augment human design capabilities and cannot often be used to create standalone designs for marketing, advertising, and branding. If you have some great ideas in mind and if you need a dependable design team that can help tweak your AI-generated images to create something commercial and useful for your brand, choose a Kimp subscription.
Register now and try Kimp for free for 7 days.