DALL-E 2 Review: Turn a Sentence into an Image in Seconds!
DALL-E was one of the first AI image generation tools on the market. It made waves for its promise of turning simple text prompts into stunning images. However, with the rise of other AI image generators such as Midjourney, how well does DALL-E 2 do? And is it worth using over the other tools current available?
John Angelo Yap
Updated September 15, 2023
Reading Time: 10 minutes
Art has always been deeply entrenched into culture.
It is a reflection of the time it was created in, from the people’s values to norms. More than that, art and media also serves as our window into the past; it’s the reason why we think of the 1940s in black-and-white or the renaissance period like Michelangelo’s Sistine Chapel painting.
Today, we live in an era where artificial intelligence is accessible. From students using ChatGPT for schoolwork to professionals automating menial tasks, there’s no denying that AI will define the next few years of technology.
Since AI is slowly turning into a necessity, it also becomes part of our culture. As such, we must ask this question:
“What does art in the age of AI look like?” That's exactly what DALL-E 2 is.
DALL-E 2 is an AI image generator that can take a simple sentence made up of a few words and turn them into a detailed, colorful image in a matter of seconds.
Over the next few minutes we’ll discuss DALL-E 2’s features and the extent of its capabilities, pricing, and competition.
How DALL-E 2 Works
DALL-E 2 is an artificial intelligence model that turns prompts into images. It’s the successor to DALL-E: a previous version of the software released by OpenAI in 2021, almost a year before ChatGPT.
To use it, simply write a descriptive prompt and wait for a few seconds as the model generates an image.
Who is DALL-E 2 For?
DALL-E 2 is an excellent tool for hobbyists and artists. It can be used to create visual concepts, frameworks, or inspiration for your next artwork.
You can also use it to make storyboards or character designs for more immersive storytelling. The use-cases of DALL-E 2 are as endless as your creativity. You can even use ChatGPT to give you ideas of what you can creative.
It’s also perfect for digital marketers who are looking to supercharge their content creation. A simple prompt and some light editing is enough to create striking marketing materials for a fraction of the price of a graphic designer.
We even use it to create the featured images across our entire site. Yes, this article too. Scroll back up & see what we chose for this one!
DALL-E 2 Features: A Rundown of AI Image Generation
There’s only one reason why people use DALL-E 2, and that’s to generate images. To accomplish this, all you have to do is write a prompt and wait for a few seconds. You can also add as much context into the prompt as you need to get optimal results. Be as detailed as possible.
The turnaround time for one prompt is usually around 30 seconds to a minute, depending on the complexity of your requirements. Each prompt will have four variations of the output, then it’s up to you to decide which one you think is best.
I've tested different types of prompts to see where it excels and where it could improve to gain a better overview of what it's actually great at doing.
General Artwork
The first thing I wanted to test was how well DALL-E 2 performs with prompts that have little to no specificity. So, the prompt I used for the first test is: “A flower vase filled with peonies.”
To nobody’s surprise, it was able to give me exactly what I needed. The results are stylized but not to the point of surrealism. The four outputs are vibrant, realistic, and different enough from each other to cover different use-cases.
With Faces
For me, the true test of a good AI generator is how well it can generate faces. When I used DALL-E during its open beta, I found myself thoroughly disappointed because it somehow only generates morphed faces which gave an “uncanny valley” feeling to the outputs. I found Midjourney to be waaaay better.
I asked the generator to give me a realistic image of a full classroom from a teacher’s POV:
I was hopeful for DALL-E but, sadly, I was let down once again. Not only were the faces morphed, but there were also some clear rendering issues with the bodies. Unfortunately, I wouldn't recommend DALL-E for generating realistic images that include people.
This comes a few weeks after OpenAI discontinued the experimental model that actually did a lot better of a job at this than the previous models. Who knows, maybe that will get officially implemented soon.
Art That Mimics Another
Aside from the contents of your prompt, you can also ask DALL-E 2 to generate images based on existing art. For instance, my next prompt was “A meadow with foxes and a tree in the middle of it during sunset, Studio Ghibli style.”
While what DALL-E produced was beautiful, it wasn’t quite what I asked for. None of the images were similar to Studio Ghibli films, and were more reminiscent of children’s books. So, I’m also docking points for this one — even if it did provide some wonderful artwork.
Logo Design
One of the handier use-cases for DALL-E is the creation of marketing materials. You can use it to create mockups, website content, and even logos.
I loaded a prompt into DALL-E for a logo for a social media content website for audiophiles. Unfortunately, I’m left unsatisfied with its run-of-the-mill and poorly-designed outputs. So, if you’re looking for a new logo, better stay clear of DALL-E 2.
It's not very good at creating text and morphing letters together. This is one of the most important things if you're looking to create any kind of logo. I'll pass for now.
Automatic Generation From An Image
DALL-E 2 can also generate artwork based on an uploaded image. I tested this functionality by uploading one of my own images as seen below.
The results were underwhelming without any editing. I was expecting more variety in the outputs but maybe a variation is really just meant to keep it super similar. Not sure what the use case on this one would be.
Image Editing
You can also edit images using DALL-E 2. For this exercise, I'll use my own photo and ask DALL-E to generate an image that puts me in a rooftop at midnight during a meteor rain. The results were, well... not what I expected:
It was a valiant effort but I'm definitely not using any of these any time soon!
Although it understood what I wanted, the images themselves have a cartoonish element.
The meteorites and the stars look straight out of clip-art. But, I couldn't discount that fact that maybe it's a fluke. I gave it another shot and asked DALL-E to put me in a farm:
Similarly, I was left dissatisfied by all variations. Definitely not good enough.
Each of them had an "uncanny valley" feeling to them, and only one actually put me in front of a farm. As far as AI-powered editing goes, DALL-E 2 has let me down for anything that needs to be photorealistic.
DALL-E 2 Pros and Cons
PROS | CONS |
|
|
Pricing
Unlike other AI image generation tools, DALL-E 2 is based purely on how much you use it.
You can currently get 115 credits for $15, with one credit equivalent to a single prompt. Each one of these prompts generate 4 images (all for the cost of a single credit).
DALL-E 2 vs. Midjourney: A Quick Comparison
You can’t talk about AI image generators without mentioning Midjourney. Since its release in 2022, tons of people have flocked over to Midjourney for its robust generation and output customizability. But, which one is the best AI image generator? I'd vote Midjourney.
Accessibility and Features
While DALL-E 2 has its own environment, you can only use Midjourney in a separate app (Discord). This limits the accessibility of Midjourney and introduces a learning curve for non-Discord users who have to get used to its ecosystem.
Beyond that, generating images is similar with both tools. You simply have to input a prompt but, in Midjourney’s case, you use the /imagine Discord command to open up where you'll put in the prompt.
What makes Midjourney less accessible to non-tech-savvy users however is, unironically, the thing that makes it so popular: customizability. Midjourney allows you to set parameters that can alter the final output. Some examples include:
- Aspect Ratio: Dimensions of the output.
- Chaos: Increases the variation of the four outputs.
- Negative Prompting: Removes certain elements in the output.
- Quality: Sets the rendering quality time.
- Style: Allows you to customize the tone of the output.
To use these, you must first learn some syntax. Nevertheless, everything that you need to learn before using Midjourney is available with their extensive and beginner-friendly documentation. If you have a few minutes to learn the personalizations, you'll get so much farther than with DALL-E.
Output
This comparison will be incomplete without seeing which AI generator performs better using a side-by-side comparison.
For this test, I’ll be feeding Midjourney the same prompts I gave DALL-E earlier and checking which one generates the better [subjective] output.
The DALL-E images will be on the left, while the Midjourney images are on the right.
So, let’s start with the vase:
I personally prefer DALL-E’s output because of how realistic it is compared to Midjourney’s output, which looks like a digital painting.
Moving on to the classroom:
To my surprise, DALL-E 2 has the better output simply because it understood the context of the prompt better. Each of Midjourney’s four generated images didn’t even include a single student in it.
Next, the meadow:
This is where Midjourney shines. Not only was it able to provide a more beautiful output, the artwork was also more accurate to the parameters I set.
Lastly, for the logo:
While I’m not a big fan of either design, Midjourney was at least able to provide a more usable logo compared to DALL-E’s subpar output.
Overall, I'd say that Midjourney is better at generating images with more parameters, while DALL-E 2 is great for generating general artwork.
The choice depends on what you're using the visual content for. I would use Midjourney more, I just wish you didn't have to go through Discord to do so.
In A Nutshell...
So, what does art in the age of AI look like?
If you were to ask me what my answer is, I’d say it’s “imperfect.”
Much like other AI generation tools, image or otherwise, DALL-E 2 is a fairly robust product that could use more time in the oven. I’m sure that in a few years DALL-E could become an indispensable tool for marketing and creative purposes — but that’s just not today.
That said, DALL-E 2’s current iteration is powerful on its own. After all, this software can turn whatever’s in your mind into actual images with just a single prompt.
There’s truly no limit to what you can do with DALL-E 2 apart from your own imagination — and that’s what makes it special. It's very affordable and can be learned really quickly.
Want to Learn Even More?
If you enjoyed this article, subscribe to our free newsletter where we share tips & tricks on how to use tech & AI to grow and optimize your business, career, and life.