Sponsored Content
If you keep up with technology and AI, you will know that when it comes to generating images, Midjourney has been everybody’s pay-to-play go-to. Now it looks like they have some competition. Generative AI is running its race, with OpenAI releasing DALL·E 3 an image generator on the 20th of September, 2023.
Ever written up an amazing blog and wanted an image to resonate with that? Ever had a really cool idea and you wanted it in a visual? Ever been too tired to create your own image and wanted it instantly? And on top of that, you wanted it to be exactly what you imagined. Well, you can do all of that with DALL·E 3.
Let’s start from the beginning. DALL·E is a text-to-image model which is developed by OpenAI using deep learning methods. We’ve seen DALL·E 2 be able to generate digital images using natural language processing and now we have DALL·E 3.
DALL·E 3 has come back bigger and better with being able to understand the nooks and cracks, more nuances and detail than ever. Using ‘prompts’ you can now easily translate your ideas into accurate digital images.
So what is the difference between the two? How is DALL·E 3 better?
Understands Context Much Better
The main difference between DALL·E 2 and DALL·E 3 is the model’s understanding of context. DALL·E 2 unfortunately had a difficult time fully understanding context even when specifically prompted, it would ignore specific words. DALL·E 3 understands context much better, providing users with the image they want.
Hand in Hand with ChatGPT
DALL·E 3 has specifically been built on ChatGPT. This allows you to use DALL·E 3 and ChatGPT hand in hand to brainstorm your ideas and better refine your prompts. When DALL·E 3 is prompted with an idea, ChatGPT will generate unique, tailored and detailed prompts for DALL·E 3 to bring to life.
If DALL·E 3 generates an image that you’re not fond of, you can ask ChatGPT to make further tweaks to get the image you want.
The Images are Yours!
Images that were created by DALL·E 2 did not belong to the user that created it. With DALL·E 3, the images that you create are all yours! This means that you do not need permission from OpenAI to reprint, sell or merchandise them. Definitely an interesting development.
Mimicking Living Artists
We won’t get into the issues surrounding why mimicking living artists is a problem – we know that you can turn ugly very quickly. Lawsuits, copyright infringement, you get what I’m trying to say here.
An OpenAI representative said that DALL·E 3 has been specifically trained to decline generating images that mimic the style of living artists. Whereas, DALL·E 2 currently can be prompted to mimic the art style of certain artists. To ensure artists are happy, OpenAI has also provided a form in which creators can opt out of having their images used to train future models.
Fake Image Generation
From what we’ve learnt about DALL·E 3, it seems like it’s an open playground. However, OpenAI is still very tight about safety around the use of all their generative AI tools. OpenAI has stated that just like DALL·E 2, DALL·E 3 has an implemented keyword and image detection filter which limits users’ ability to generate harmful, violent and sexual content. We’ve already seen this happen with Midjourney when it generated fake images of Donald Trump getting arrested.
Have a look at the below image of the difference between DALL·E 2 and DALL·E 3 on their output on generating an image using the prompt ‘An expressive oil painting of a basketball player dunking, depicted as an explosion of a nebula’.
Image by OpenAI
DALL·E 3 has very recently been rolled out to ChatGPT Pro, with availability coming soon to OpenAI APIs and Labs as well.
To use DALL·E 3 from ChatGPT Pro, with the convenience of interacting with the service via the familiar chat interface, simply head over to the ChatGPT website and from the ChatGPT-4 menu option select “DALL·E 3 (Beta).”
Screenshot from ChatGPT website
At this point, all you have to do is interact with ChatGPT in the same way you would otherwise.
Create an image of a mountainous winter scene, with a cabin and some goats
And here’s what DALL·E 3 generates and outputs right inside he ChatGPT interface:
Image by Author using DALL·E 3 (click to enlarge)
It’s that easy. ChatGPT takes care of engineering useful prompts for DALL·E to use, making the system far more approachable than some of the other options out there which require clever prompt engineering to get their best results.
And that’s DALL·E 3 as it is at the moment. What does this mean for other AI image-generator competitors such as Midjourney and StabilityAI?
Let us know your thoughts in the comments below.
Nisha Arya is a Data Scientist and Freelance Technical Writer. She is particularly interested in providing Data Science career advice or tutorials and theory based knowledge around Data Science. She also wishes to explore the different ways Artificial Intelligence is/can benefit the longevity of human life. A keen learner, seeking to broaden her tech knowledge and writing skills, whilst helping guide others.
