Using OpenAI’s latest AI image generator is easy! This is not great news for leading competitors Midjourney, and Stable Diffusion from Stability AI. AI-generated images have ushered in a new era digital art, with proponents and opposition on either side. To keep yourself ahead, here’s how to use DALL-E 3 (Stylized DALL·E 3).
Where can you access OpenAI’s latest AI model?
ChatGPT Plus and ChatGPT Enterprise subscribers can access DALL-E 3 via ChatGPT. Simply enable DALL-E 3 and type a text prompt as usual — if the AI chatbot detects that you would want an image output, it will generate one. The AI image generator is also accessible through the OpenAI API or via OpenAI Labs (a subdomain of the OpenAI website).
Eligible customers are able to use DALL·E 3 via ChatGPT (app & web browser), Bing’s AI Image Creator, and Microsoft Designer.
DALL·E 3 is now in research preview, and will be available to ChatGPT Plus and Enterprise customers in October, via the API and in Labs later this fall.
OpenAI.com
DALL·E 2 is available for anyone to use, via desktop or mobile, and can do much the same thing as DALL·E 3 (albeit to a lesser degree of quality). This older model of AI art generator also has a plugin, which allowed users to generate DALL-E 2 images — now redundant thanks to the native integration of DALL·E 3. All three versions of DALL-E were created (or co-created) by Aditya Ramesh, Head of DALL·E at OpenAI.
Essential AI Tools
Content Guardian – AI Content Checker – One-click, Eight Checks
Originality AI detector
Jasper AI
WordAI
Copy.ai
How to use DALL·E 3
Now publicly released, use of DALL·E 3 is limited to paying ChatGPT Plus and ChatGPT Enterprise subscribers, and is not available with the free version of ChatGPT. This will likely make ChatGPT Plus worth it for you, especially if you’re an active ChatGPT and Midjourney user.
The new tool will combine a diffusion-based neural network with a large language model (LLM). The latter with take text prompt in natural language and, through a process of natural language processing (NLP), instruct the imagery generation of the diffusion model. Thankfully, you don’t really need to know how it works in order to use it.
You will need a paid subscription, however. In addition to this recurring monthly cost, you’ll be required to buy credits in the top-right corner of OpenAI Labs.
Is DALL·E better than Midjourney?
DALL·E 3 is better than DALL·E 2 in multiple ways. This includes aesthetic quality, usability, and even text-in-image generation accuracy. While OpenAI’s new AI image generator is unlikely to beat Midjourney in aesthetic quality, the latter two considerations are really where it shines.
For comparison, here’s Midjourney v5.1 executing the exact same text prompt:
Undeniably, Midjourney images are more usable than DALL·E 2. The latest iteration can edit images and make higher quality logos than its predecessor, but still can’t generate GIFs. However, DALL·E 3 is comparable in quality to Midjourney 5.1, as shown below (not with an identical prompt).
In addition to aesthetic quality, DALL·E 3 far exceeds the competition at realistic text generation and font legibility in AI-generated art, becoming the best AI model for text-in-image generation from any big tech company. As a result, Midjourney will have to release new competitive features in order to continue being worth it alongside alternative AI art generators.
Leaked DALL·E 3 AI images
Early examples of DALL·E 3 AI-generated images were leaked on Discord earlier this year. The leaker reportedly used the following prompt, and received this corresponding AI images. Following the release of official images, the below ‘leaks’ resemble the older model DALL·E 2 far more than the one the leaker claimed to have access to.
“painting of a pink jester giving a high five to a panda while in a cycling competition. The bikes are made of cheese and the ground is very muddy. They are driving in a foggy forest. The panda is angry.”
Mashable.com
Steve’s opinion
DALL-E 3 is, aesthetically, about as good as Midjourney. Lines are clean, concepts well executed, and it’s as capable with a very wide range of art styles as its key competitor. Ultimately, DALL-E generated images are an accurate portrayal of the text prompt. However, the one downside that keeps me subscribed to Midjourney is the restrictive censorship of DALL-E.
There’s a line to be drawn with regards to portrayal of sensitive subjects, and I don’t claim to know where that line is – but with many users reporting via social media that it’s hardly possible to generate any image wherein a character is holding anything that could be described as a weapon, it’s clear that a PG 13 image generator is not what users really want.
You might notice these personalised ‘final thoughts’ on new articles. We’re introducing more of a personal touch here at PC Guide, and if you’re a fan then please let us know via Twitter @pcguidedotcom.
