Generative AI will replace stock libraries... or will it?

With new tech, it’s very easy to signal the end for methods that have been the norm for years. But the reality is, replacing tried and tested methods (even if they are not perfect) is not as easy as it appears on the surface. Let's explore the potential and pitfalls of turning to AI to generate images for your campaigns...

Anyone who’s worked in a creative team knows the pattern when it comes to advertising campaigns. Particularly when it’s a regular quick turnaround below-the-line campaign. The concept is created and as part of that invariably requires spending hours on various stock library sites (depending on budget) trying to find images that fit the campaign narrative.

Now let me say this… Stock libraries are great! And the uplift in quality and diversity is noticeable. Some great work has been done over the last few years to provide more regional and inter-continental options in imagery. There’s still work to do, but they have served us well.

But how often have I heard the phrase, “It’s almost right” and that’s the issue when it comes to stock? It’s very rarely been shot for your specific brief. So you are always making concessions.

Bigger budgets and campaigns can fund photoshoots, which are a great way of being able to control the narrative of your campaign, but in so many cases at the time of the shoot, no one’s considering the photo might be used on a skyscraper, and a 728x90 display ad in three months.

This is where I see AI becoming a hugely influential tool in the armory of campaigns!

Take the following shot, typical of what you can find in any stock library, or shoot, framed for traditional advertising without scope for use out with a 96-sheet or double-page spread.

But if you need to roll that imagery out over display and social (which you should be doing if you are truly running an integrated campaign) in the past you would be limited in your options.

The image above shows the limited crops that are available to you and stops the image from being a true reflection of the campaign vision.

Generative AI enters stage left…

With Photoshop’s generative AI tools, you can add extra space to the photo and add new info to it in minutes.

The result… perfectly formatted images for your display assets! Bringing much more of the photo into the viewable space, and arguably a large area to get text, messages and CTA’s in the space too.

But that’s just scraping the surface. What about going one step further?

We recently did a client campaign to use their downtime while commuting to consider their financial future.

Hours, and I mean HOURS were spent trawling image libraries for appropriate imagery, but wouldn’t it be great to gain more control over your environment with Generative AI, giving you the scope to control elements of the environment and subject matter to better suit the brief?

Of course, to get the best results a thorough experience of prompting and engineering the correct prompts will be key to the level of success of this approach. But in terms of capabilities, AI tools are more than capable of generating some striking results.

The top shot is a stock shot we used in the campaign. The middle two are generated by MidJourney and the bottom is from Adobe Firefly.

Is it viable yet?

There are limitations at the moment around resolution and size, less of an issue in the digital space, but when it comes to OOH or print MidJourney Dall.E, NightCafe and Stable Diffusion don’t provide high enough quality imagery out of the box, but there are some commercial upscaling products which will provide some good results.

What are the legal implications?

The quick answer is, we don’t know yet.

A big concern at the moment is the lack of clarity around the legal standpoint, there is already a copyright lawsuit in the US against Stability AI and Midjourney that claims that these organisations have infringed the rights of “millions of artists” by training their AI tools on five billion images scraped from the web “without the consent of the original artists.” Until this is resolved it’s a potential minefield when it comes to using AI-generated imagery.

And world-renowned fantasy artist Greg Rutkowski has had his name used in the generation of AI on a huge scale with no permission. "It's really hard to tell whether this will change the whole industry to the point where human artists will be obsolete. I think my work and future are under a huge question mark."

It’s worth noting that Adobe’s Firefly gets around this issue by limiting its image bank to images available through Adobe Stock. Good from a legal clarity standpoint, but not so good when it comes to variety and choice as there’s a limited pool of references. That said, running the same prompts I used in MidJourney through Firefly delivered results arguably closer to the original Shutterstock image I based my test on, but there is more work to do and it doesn’t have the polish and quality MidJourney does yet.

So are stock libraries on the way out?

Absolutely not, over the years stock libraries have become very adept and changing their business model (remember buying stock images on a DVD?)

I see them not just serving a purpose for the foreseeable future, but driving a sustainable generative AI model for businesses wanting bespoke imagery for their campaigns, but without the time to become professional prompt engineers. Already they are introducing AI tools into their functionality and I see an evolution of the current stock library marketplace to one that provides custom imagery to clients rather than generalist imagery that might fit your campaign.

So where do we stand?

With the landscape changing so rapidly and the proliferation of the number of tools being released on an almost daily basis, time is required for the whole generative AI space to settle, and allow the most popular tools to establish themselves. No doubt the AI tools that win out will be the ones that can resolve the legal ramifications and provide clarity to users. Looking at recent history and the digitisation and streaming of the music industry suggests it’s very possible but what that may look like, we don’t know yet.

Are you considering using AI to generate images for your marketing activity yet? If so what are your views on how the landscape is going to evolve over the next months and years?

Header Image: https://futurism.com/the-byte/ai-awful-generating-pictures-human-hands