Exploring AI tools for the holidays: DALL-E

In previous blog articles, we’ve explored the ways in which AI tools can now assist us with creating incredible and realistic images, such as Midjourney, or professional looking designs from scratch (like Canva does). But there’s more of those out there, with varying levels of accessibility and capability - one of which is our main subject for this article.

First Iteration of DALL-E

In terms of text-to-image generators that use generative AI to function (i.e. the most common form of AI image generators known today), DALL-E is known to be one of the first to exist. Created by OpenAI in 2021, the first version of DALL-E is sorely lacking by today’s standards, yet back then (just a mere two years ago!) It caused a huge sensation because it seemed able to create art that is both unique and yet familiar at the same time.

In fact, it’s possible that the ‘uncanny valley’ effect achieved in DALL-E’s first outputs inspired its own name: Salvador Dali, a famous surrealist painter, was well known for dreamlike illustrations and depictions in the world of fine art.

DALL-E 2

Just one year after its debut, OpenAI revealed the next generation of DALL-E to the public. Instead of hazy washes of colour that hinted at objects or people, DALL-E 2 now promised better quality images, more realistic depictions, and consistency in specific art styles when instructed to do so.

At the time of writing, DALL-E 2 is still the official standalone version of the image generator available to the public, which can be accessed here. Just like Midjourney however, DALL-E 2 does require an upfront payment of at least $15 USD, which nets you 115 ‘DALL-E Credits’ - each image generation request to DALL-E uses one DALL-E Credit, as does each edit request for an existing image.

With all that said, there exists a newer, better version of DALL-E that’s also available…

DALL-E 3

On 19th October 2023, OpenAI released DALL-E 3: currently the latest iteration of the image generator series. Designed to be integrated directly into ChatGPT, users are able to hold human-like conversations with DALL-E for the first time - instructing, directing, and collaborating with the AI to create desired images and illustrations.

But this improvement in user experience isn’t all that DALL-E 3 has to offer. Featuring even better accuracy in image output and a higher understanding of what users want in their requests, DALL-E 3 is able to tease out the exact requirements in each task much better than its predecessors and translate that into highly impressive artwork, regardless of style.

But there are two major catches for DALL-E 3. First, it’s only usable within ChatGPT, so you’ll need to sign up for an account in order to use it. Second, it’s only available to ChatGPT Plus subscriptions, which costs $20 USD a month, so you’ll need to pay for it first before you can try it.

This might sound like a dealbreaker to most, but considering that a ChatGPT Plus subscription also grants access to GPT-4, OpenAI’s latest AI model for ChatGPT, as well as a slew of advanced features, you’ll be surprised to know that the subscription was so popular that OpenAI had to pause new signups for several months in late 2023 because they just couldn’t keep up with demand.

Future of DALL-E & Other AI Image Generators

The race towards advanced AI systems is well underway, but there’s no clear winner yet. As an image generator, DALL-E faces stiff competition from established names like Midjourney and Stable Diffusion, with traditional tech giants like Photoshop and Getty Images only just leaping into the fray.

No matter who wins this race however, one thing is for sure: generative AI-produced images are here to stay, and they’re steadily becoming much better at it than we have ever thought possible.