Words-to-image AI creator tool DALL·E is now available to all, without a waiting list.
This is big news, and could have huge implications for artists, marketers and the wider media.
If none of that first sentence makes sense to you, I’ll briefly explain.
What is DALL·E and why is it making headlines?
DALL·E , created by artificial intelligence company OpenAI, is a portmanteau of Salvador ‘Dali’ and Pixar’s ‘WALL-E’.
It is one of a number of AI image generators that takes user text prompts and uses machine learning to create photo-realistic images that have never existed before.
More than 1.5M users are now actively creating over 2M images a day with DALL·E—from artists and creative directors to authors and architects—with over 100K users sharing their creations and feedback in our Discord community.
Since January 2021, OpenAI has made quick work of getting to this point, taking data points to return images based on millions of existing pictures and captions to create images that match that description.
Tech site Pocket-Lint called it ‘the best image AI system [we’ve] seen’, in this post looking at ‘incredible images’ created by text prompts.
To see other examples, I’d suggest this Subreddit, or OpenAI’s own Instagram and Twitter accounts.
From examples like the above to examples OpenAI give you when using the tool like this cute little Pixar-esque thing, it’s easy to get excited.
Why should the PR industry and other marketers care?
As with any new and exciting tool, it’s going to be of interest to brands and agencies, while we all work out what the potential is – and limitations are.
Heinz made (mostly marketing trade) headlines in August when it claimed that ‘even computers know that ketchup is Heinz’.
Heinz didn’t specify that it’d used DALL-E, but the brand was definitely capitalising on the hype to press its core marketing message.
Even then, users were quick to say that they didn’t get the same image response, and today; now it’s open to all, you can see that it’s just not the case.
Here’s one I made earlier:
Commercial restrictions?
If you’re wondering about the legal and commercial restrictions, OpenAI has been clear in this content policy answer.
Users have full rights to commercialize the images you create with DALL·E, so long as you follow our content policy and terms of use. These rights include rights to reprint, sell, and merchandise the images.
It should be said, not everybody thinks this is A Good Thing, as OpenAI’s comment above doesn’t take into consideration how artists’ work incorporated without consent is used in its dataset, for instance.
How could PR and marketing professionals use DALL·E?
There are a few use cases I can see DALL·E being of use to PR pros, advertisers, social media managers and the wider media.
1. Creating quick, rights-free imagery to complement articles and blogs
There’s a lot in the media right now about the cost of living.
Not one of the four photos below existed before I created them, but you can see how each of them could very easily sit beside an article about the cost of living.
Of note: Getty Images has banned people from uploading AI-created images to its library. You can see how stock imagery libraries might be hit by DALL·E, but this move has more to do with copyright concerns. Getty said: ‘there are open questions with respect to the copyright of outputs from these models and there are unaddressed rights issues with respect to the underlying imagery and metadata used to train these models’
2. Creating assets to complement a PR campaign idea
Here’s where you can let your imagination fly. You know those ‘here’s what Princess Jasmine would look like as a real person’-type stories?
Because DALL·E uses existing data-sets (with an attempt to eradicate racial bias, as mentioned in the section below), we can pull on pop-culture and more in our prompts.
Here, I’ve asked it to imagine Uncharted’s Nathan Drake as an old man. The results are certainly up there quality-wise with what I’ve seen in other artist-completed briefs and branded character re-imaginings, popular across digital PR campaigns.
Now, that poses a question in itself, potentially threatening the livelihoods of artists and designers, as noted and touched on in AI artist and expert shardcore’s response at the end of this post. I’m not necessarily advocating for it, but considering use cases while we all wrap our heads around it.
3. Social media and website imagery
Need a photo of a group of friends eating takeaway while laughing, but your company doesn’t want to go to the expense of putting the shoot together? Or need a photo of a dog enjoying its dinner?
Again – these photos do not feature a single real person, or a real dog (side note: I’m sure we’ve all seen This Person Does Not Exist and the like?).
4. Presentation and pitch asset design
Very often, when pitching, clients and potential clients want to see what a creative idea MIGHT look like.
Here’s where DALL·E can help us visualise our creative thinking, which can often unlock budgets.
I’m obviously being daft with the below trope, but you can see what I mean!
5. Product or service concept imagery
You’ve seen the type of story. ‘X fast food brand teases drone delivery!’, or ‘limited edition range of Y product will be released’ – complemented by concept imagery.
Personally, I’d love my food to be delivered by a Frankie Zapata/Green Goblin-inspired delivery driver!
6. Brand collaborations
Every year, Evian has another famous designer reimagine its bottles in a limited edition designer range. Here’s what DALL·E thinks it’d look like if created by Louis Vuitton.
What would a collaboration between one of your clients and another brand a bit left-field look like? Can DALL·E help you sell it in to each party?
Can you think of any other use-cases? Tweet me and I’ll add with credit!
And the downsides?
It’s easy to imagine how DALL·E could be misused.
OpenAI has attempted to get ahead of it by addressing the issue directly. Whether this is something users can get around is to be seen.
In the past months, we’ve made our filters more robust at rejecting attempts to generate sexual, violent and other content that violates our content policy and built new detection and response techniques to stop misuse.
There are also social issues to be considered when considering AI-generated images.
As reported in The Verge’s piece ‘DALL-E is available for anyone to use immediately‘:
DALL-E invisibly inserts phrases like “Black man” and “Asian woman” into user prompts that do not specify gender or ethnicity in order to nudge the system away from generating images of white people. (OpenAI confirmed to The Verge that it uses this method.) This does mitigate bias in DALL-E’s output, but some users have noted it also creates unwanted imagery that doesn’t match their instructions.
Taking a broader view of the news, experienced communicator Matt Muir, creator of weekly newsletter Web Curios, said,
“If you’re a junior designer (or the sort of freelancer who gets paid to churn out fast-turnaround pitch visualisations) you’re f***ed, or at the very least should have a section on ‘working with AI’ in your portfolio.
“It also means brands need to be careful just rinsing an artists’ style in an AI image because they can – someone will get sued quite hard at some point when lawyers get their heads round this properly. At the moment it’s very grey, and you can have some short-term edgy fun ripping off known visual vibes, but I can see it getting messy and litigious quite quickly.”
And, as teed up earlier in this post, AI artist and expert shardcore said:
“AI “art” generators are all the rage. Various iterations (with variable skills) have been arriving almost weekly over the last 12 months or so. Dalle-2 is one of the ‘big boys’, coming as it does from OpenAI (who run a very slick PR department). The public release is significant, opening up access to anyone.
“As ever in the world of AI, the legal system is completely unprepared for what’s occurring – the wholesale ingestion and assimilation of millions of (potentially copyrighted) images from the web, and a machine that can spit out compelling replicas of famous people, branded characters, logos and also artistic styles poses enormous ethical (and inevitably legal) questions (watch out for a slew of court cases in the coming months.)
“While we wait for the law to catch up, the first rush of horror has come from practising artists, watching as their style is stolen (if they’re in the dataset) and livelihoods threatened by a machine that allows anyone to type in some words and get an infinite number of pictures in return.
“Many will argue that this announcement from OpenAI ‘democratises’ the production of images, though I fear, as with all radical technological advances, it will result in a massive “disruption” and “rebalancing” of ’the creative economy’, and that’s rarely much fun to actually live through.
“These technologies will, of course, become assimilated into our lives – clickbait articles about “the death of art” are as premature as those that accompanied the invention of photography almost 200 years ago.
“We still need artists, and art will adapt to accommodate these new tools, it just might be a little bumpy along the way.”