When new technology emerges, it always feels like magic. I remember when GPT-3 started making the news. I waited for months to get access and was absolutely blown away when I typed a sentence, clicked submit, and the AI just ran with it. It could write an entire article or research paper, even a novel. Now, text generation programs are easily accessible and cheap.
The new AI kid on the block is image generation. Type in a text prompt, and within seconds, a grid of images appears. It’s like magic. Want to see Chewbacca working an office job? Got it. Want to see teddy bears mixing chemicals in a laboratory? Wait no longer. Want to see your celebrity crush nake…. wait, no. That’s not allowed.
These AI image generation programs have been trained on God only knows how many pics. And they’ve put some parameters in place to make sure they are not used in illegal, harmful, or inappropriate ways. Thank you, AI overlords.
If you are interested in how the generative age is evolving, the power of these image generation tools, and the unique aspects of these two programs, I’ll show my side-by-side comparisons of DALL-E 2 and Midjourney and point out some of their key differences. Now, on to the images!
Let’s start with a simple prompt that most people can compare to an image in their minds. I typed, “Brad Pitt” in the prompt engines. Here are the results:
As you can see, there are a few differences here. DALL-E 2 is in general more photorealistic, while Midjourney is more artistic. This stays true throughout and seems consistent with each company’s stated goals. Midjourney has focused on building something that closely mimics artistic styles and empowers the user to create new forms of art. DALL-E 2 is better at photorealism. I suspect this has much…