As a professional colouring-in person I hate it, and it will become boring very quickly, but I know my customers will start asking for this style soon as it's all over social media! 😱
So I feel I should know how it's created so they can create their own 'designs' and send it to me for print.
Is it ChatGPT, Canva or AN Other? I've looked on Canva but can't see any templates for this style.
Microsoft Publisher? 😂
IANAC, but they don’t look anything particularly special to me
That looks similar to those profile pics I'm seeing all.over Facebook at the moment. I think it's ChatGPT folk are using.
It's certainly AI. ChatGPT is as plausible as any.
Google Lens says it could be almost any AI tool as long as you use a text prompts describing a "vintage, minimalist, text-only poster with a green background" to generate design
As a designer and illustrator there aren't the words in the English language to truly express how much I utterly despise the proliferation of this style of AI generated visual diarrhea.
It certainly sends a message. Unfortunately that message is that I'm so lazy, devoid of ideas and hold potential customers in such contempt that I couldn't be arsed to do anything more taxing than ask ChatGPT to generate an image for me
Prompt to Google Gemini: Nanobanana, make an advertisement for a bicycle shop in the style of this (link to image)
Result:
As others have said it's AI - it looks horrid I hate it.
Our local RSCPA started doing them to advertise dogs for adoption & people started calling them out in the comments.
Totally agree binners. I have no skin in this game other than as a consumer and that is vile, homogenised slop of the worst kind. I would actively avoid any company pushing out that sh!t.
Better get used to it sadly! Though if its anything like the coding models, the quality of the output will improve at least.
Better get used to it sadly! Though if its anything like the coding models, the quality of the output will improve at least.
Will it though? When the models run out of human creativity to steal they will be left feeding off each other.
Better get used to it sadly!
I'm betting that if people aren't bored with everything looking exactly the same already - and god knows, I am - they will be in about a week
I would actively avoid any company pushing out that sh!t.
I have zero intention of using it myself. But I run a print-shop and can't afford to turn work away because I don't like the style or how it's created. I've printed many beautiful leaflets full of rainbow word-art over the years!
I do very little design work now* - Canva is used by loads of my customers and they just send me links to those files.
(*a lot more money in print).
I hope you're right, but i'd take that bet all the same 😔
Sadly, it's AI slop. Looks utterly terrible as well. They're all totally identikit. I'd expect it's mostly ChatGPT, as that's AI to the vast majority of the public.
There'll always been a premium for real human work, but a smaller number of clients willing to pay it. I fear that designers / coders / writers are now like bespoke tailors. Sure, they do a better job, but most folks will go for the £200 suit that sort of fits rather than the £4,000 one that actually does.
Model makers aren't just stealing, they are also employing people to generate new stuff for the models to ingest. Not me trying to justify it or anything, but they ran out of new material to train on a long while ago, and the models still improve.
To be honest, I don’t even care if a human’s work is “better”, or not… even if it is “worse” it will still connect with people better than this kind of slop.
I guess the breakthrough moment will be when it takes an expert to assess if it is AI generated. It's not there yet with this kind of stuff - it's instantly recognisable by most of us.
I feel sorry for any younger designers or writers or anyone in a creative field, trying to get a foothold in these industries. I'm lucky in that I'm in an area where I work with clients who want bespoke, well-designed comms and wouldn't touch this kind of lowest-common-denominator shite with a barge-pole.
I know because I've seen it though, that there's already an attitude amongst a lot of management of "can you not just use AI to do it?", and a lack of concern for the quality of the output as long as its not costing them anything and they can tick another thing off their to do list
Isn't this a good thing for professional graphic artists though? Surely your original work will stand out so starkly against this garbage that discerning customers see the advantage it brings.
It's not there yet with this kind of stuff - it's instantly recognisable by most of us.
But it's come a long way, very quickly. And as I read as I read once, "AI today is the worst it will ever be."
I doubt this displaces good designers, it just makes slightly less horrific DIY efforts in paint/word with comic sans/word art 1990’s clip art.
if you were a designer who had a dozen templates he quickly adapted for the lower end of market it might be a bit worrying, but anyone who’s ever tried to get AI to “just like that last image but with a small change” will know talking to a human would be much easier - eg. If thols liked that image but thought - I’m trying to appeal to women I bet swapping JUST the rider for a female is not easy, or saying “can we put a family there”. My experience with ChatGPT is you can something totally different - it might be slightly better. Then you try saying, ok I like that, but the text was better on the first one and you get something different again. I did this for a sports event and it was so infuriating I swore never to try it again. No doubt it will get better at understanding me but it doesn’t really understand the elements of the picture so it’s really hard to work with and i don’t think most of the output is editable in a graphics package (eg to correct a typo, remove a word you don’t like, change a price etc). It can’t be far off that though.
Isn't this a good thing for professional graphic artists though? Surely your original work will stand out so starkly against this garbage that discerning customers see the advantage it brings.
That's why I specifically mentioned younger designers. When you start out in the industry you wouldn't be working on the kind of things I am nowadays (I'm an old duffer, who's been at it for decades). It's like any other job I suppose. You start off doing the more mundane day to day stuff while you learn what you're doing and work your way up to doing more and more involved work
Unfortunately this is exactly the area where companies are thinking 'we can get AI to do that' or "Sarah in admin can knock something up in ChatGPT" . It'll be rubbish but unfortunately I think it's going to seriously impact the opportunities for younger designers. And if they don't get those opportunities then where do the more experienced designers come from, once that avenue is closed off?
I'm not anti-AI. I use it all the time. The photoshop AI is great for saving the donkeywork of editing multiple images, for example. But I worry that some people now see it as a replacement for highly skilled designers and the end result is a flood of this type of crap, which I absolutely loathe (purely from an aesthetic standpoint)
Surely your original work will stand out so starkly against this garbage that discerning customers see the advantage it brings.
One of the reasons you employ a designer, is that you want/need them to be discerning for you.
I bet swapping JUST the rider for a female is not easy
It is though. Asking the LLM to do that will result in a "good enough" result very quickly, rather than being charged a day rate for a designer to make the amend and do a better job. Here's why the race to the bottom is unavoidable... for many people cheap and fast wins... and without a designer "in the way" the client gets to satisfy their own tastes, discerning or not. And the world gets more slop.
I guess the breakthrough moment will be when it takes an expert to assess if it is AI generated. It's not there yet with this kind of stuff - it's instantly recognisable by most of us.
I see that style of image all over the place now and whether it be on YouTube, web pages or as headers for online content I actively avoid opening the page or YouTube video, utter ****ingbtripe that boils mynpiss
20 years ago, reports were filled with clipart. Now we have AI generated images. I pay for the magazine to read the words, not the pictures.
It is though. Asking the LLM to do that will result in a "good enough" result very quickly, rather than being charged a day rate for a designer to make the amend and do a better job.
I've not tried for over a year now but it was way harder than I expected it to be. Now Grok can apparently nudify people so perhaps its got there. What I found in ChatGPT4 was it could produce something like the spoke n shift advert above - but whenever you asked it for a tweak it reproduced the image with other thinks that might be wrong, or at least not to your liking. It honestly would be interesting if @thols2 was to say "that is nice, but I'd like it to show a female rider" and see if it just swaps the person. Similar "can you produce it without the creased paper effect" or changing Open Daily to Open Tue-Sun, would suddenly change bits of the image you never asked it to touch.
It might be that "Nanobanana" is better at this - but then you need to know the right tool to use. Nanobanana is based on Gemini - which keeps offering to do stuff for me in google slides - but seems to not understand their own product and wants to turn vector based "slide" composition into fancy, but uneditable png's! Of course again it could be user competence. I think if you actually want a common brand/theme then AI is still someway off - especially across multiple tools, but will of course catch up.
I think I'm mostly in agreement with Binners here, but I think there is a use case for it. It shouldn't - but probably will - displace human artists, but you don't always need professional art. The Clipart analogy is a good one; Clipart was garbage, it used to be annoyingly ubiquitous garbage, but it's better than I could do in MS Paint and if I'm knocking out a poster for 3-year old's birthday party then it's Good Enough.
I was going to say that, of course, it's easy for me to be dismissive because AI art isn't a threat to my livelihood. But AI more generally is. It's got its tendrils into software development, report writing, all manner of technology areas. I predict that it's going to come back and bite us at some point, because it's generating code which is functional and that's fine for a dirty script you want to run once and then bin, but not so great for production systems which need to be robust. The bad guy hackers are going to have a field day (but that's OK because we now have AI doing penetration testing...)
As a designer and illustrator there aren't the words in the English language to truly express how much I utterly despise the proliferation of this style of AI generated visual diarrhea.
It certainly sends a message. Unfortunately that message is that I'm so lazy, devoid of ideas and hold potential customers in such contempt that I couldn't be arsed to do anything more taxing than ask ChatGPT to generate an image for me
I certainly know where you’re coming from, and understand completely, however, this isn’t much different to the formula used back when I first started in print and publishing, where people used books of standard layouts which could have different text added using Letraset or typeset with whatever fancy font was handy, and copyright free illustrations, because the client wanted something cheap and cheerful to push through letterboxes or hand out at events.
The great majority would end up in the recycling/landfill/on the ground, so quality was irrelevant.
This is just a bit better quality version of cheap advertising slop that’s been coming through your letterbox for decades, for double glazing, roofing, tree and hedge cutting, local takeaways, etc. I’m no longer working, but if I still had a computer that could handle that sort of stuff, I’d probably use AI graphics, just because the nature of that sort of thing really doesn’t require much effort expended on it.
If you could charge £50 for five minutes work, or even double that, and the client is happy, then I’d consider that a win and pocket the money; after all, it’s not fine art they’re paying for, just cheap recycling material. 🤷🏼♂️
which is functional and that's fine for a dirty script you want to run once and then bin, but not so great for production systems which need to be robust.
Yeah its quite a bit past that now. Not just for script kiddies. Still requires supervision, and quite a lot of expertise to prompt it in small steps but with every model release each spin on the LLM wheel requires less work to knock it into shape. Coding is pretty rapidly dissapearing as a craft, at least if what you are building is heavily represented in the training set (eg, all web and mobile dev, yes even distributed systems). Technologists are increasingly looking more like some combination of BA + product owner (to write prompts) and Software tester (to validate the outputs).
Sure, they do a better job, but most folks will go for the £200 suit that sort of fits rather than the £4,000 one that actually does.
And then the skilled folks will be left trying to pick up the pieces after the crummy vibe-coded whatever reveals the client’s API keys and acts like a sieve in keeping their information secure.
And then the skilled folks will be left trying to pick up the pieces after the crummy vibe-coded whatever reveals the client’s API keys and acts like a sieve in keeping their information secure.
Hooray your job is safe I guess? Anyway, as I said above, models and tooling (eg codex, opencode etc) have moved well past the early versions that were hillariously bad. If you are basing your assessments on GPT on its own from 18 months ago then you'll be well off the mark. With good tooling, prompting and review, LLM's produce code thats more than good enough. Faster and cheaper than a human can. And every ~6months the amount of review thats required decreases. The job has changed. LLM's (and other gen AI) are not going away.
Some cope that has some credibilty:
Models will stop improving and hit the wall - fair enough, but no sign of it yet
VC bubble hyperscaling puts the customer cost well below what it actaully costs to run these things - maaaaaybe, but DeepSeek et al have been able to make things for a fraction of the cost of the bigger SV outfits. Models are looking more like commodities now.
This all just an epic bubble and will all end in tears - probably, but that doesn't make the tech any less useful.
Better get used to it sadly!
I'm betting that if people aren't bored with everything looking exactly the same already - and god knows, I am - they will be in about a week
Yep, I think I saw one and thought AI shit then within a week they’re everywhere.
I've not tried for over a year now but it was way harder than I expected it to be.
The models I use at work (coder) are completely different league to those a year ago.
It's not just the models though that are evolving, how they are used is evolving too. If you are just throwing a sentence in as a prompt then you might get something ok, but there is a lot of research in to how to structure the data you give to the models to produce a good output, e.g. https://github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/ace-fca.md is something I've read and tried recently with astonishing results.
I use AI and do agentic coding basically full time now, a year ago I was a sceptic. Yesterday I wrote (with my robot friend) a complex prototype across multiple huge codebases for something in a few hours that would have taken weeks had I done it by hand. That prototype and the data will produce reports and plans which I'll give back to the agent and write something production ready. it'll take a fraction of the time and effort that it would have taken a year ago.
I don't know why more people aren't worried about the impact. Maybe it is particularly effective at my job and I see it first hand so have a real day to day feeling of "shit I'm going to be redundant soon", whereas it might not have clicked for everyone else yet, they're just happy churning out posters on chatgpt.
And then the skilled folks will be left trying to pick up the pieces after the crummy vibe-coded whatever reveals the client’s API keys and acts like a sieve in keeping their information secure.
Hooray your job is safe I guess? Anyway, as I said above, models and tooling (eg codex, opencode etc) have moved well past the early versions that were hillariously bad. If you are basing your assessments on GPT on its own from 18 months ago then you'll be well off the mark. With good tooling, prompting and review, LLM's produce code thats more than good enough. Faster and cheaper than a human can. And every ~6months the amount of review thats required decreases. The job has changed. LLM's (and other gen AI) are not going away.
I was using agent mode in vscode to harvest data from the k8s stuff we had,it’s impressive in its writing throw away code,literally minutes instead of an hour or so digging up the syntax and twiddling.
It would do all the commands to get the names of the clusters and do the context switching.
I’ve also used it to trawl over hundreds of config files and all jobs that are pretty basic.
I actually do know the syntax or where to get it and what I was doing so when it went slightly iffy I could get it back on track but you could literally replace me with a junior programmer and save yourself plenty of buckeroos.
The issue is when the user is unable to validate that the results are viable,I’ve seen other non programmers do some iffy stuff and it was worrying to see the amount of info regurgitated from the AI and feed up the chain as advice to management. The AI would always hone in on a single error and make it the single root cause when experience would make you wary.
The seduction 🙂
if I still worked I’d probably be tailoring my own LLM to rewrite the legacy back ends in a style I wanted.
If your career is in programming and your not all over this shit it’s going to be all over you 🙂
Its just another tool,terribly hyped but now it’s out the box, it’s not going back in.
@chambord Haven't you just explained why your job isn't redundant though; the skill of structuring the prompt that produces a good output is going to be in demand isn't it?
I've not tried for over a year now but it was way harder than I expected it to be.
The models I use at work (coder) are completely different league to those a year ago.
It's not just the models though that are evolving, how they are used is evolving too. If you are just throwing a sentence in as a prompt then you might get something ok, but there is a lot of research in to how to structure the data you give to the models to produce a good output, e.g. https://github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/ace-fca.md is something I've read and tried recently with astonishing results.
I use AI and do agentic coding basically full time now, a year ago I was a sceptic. Yesterday I wrote (with my robot friend) a complex prototype across multiple huge codebases for something in a few hours that would have taken weeks had I done it by hand. That prototype and the data will produce reports and plans which I'll give back to the agent and write something production ready. it'll take a fraction of the time and effort that it would have taken a year ago.
I don't know why more people aren't worried about the impact. Maybe it is particularly effective at my job and I see it first hand so have a real day to day feeling of "shit I'm going to be redundant soon", whereas it might not have clicked for everyone else yet, they're just happy churning out posters on chatgpt.
Most definitely, I think a lot of programmers have been hiding behind bloated convoluted frameworks for many years and AI just blows the doors off that.
The issue I have is there’s no Intelligence in it , you’ll get a huge load of tat generated when you could do something more elegant and simpler in HTMX but AI doesn’t innovate ,it just regurgitates.
@chambord Haven't you just explained why your job isn't redundant though; the skill of structuring the prompt that produces a good output is going to be in demand isn't it?
Kind of agree for now, but the overall impact will be a contraction of the jobs market, less jobs for younger/junior engineers. Eventually I suspect I will no longer be required because a more senior/experienced engineer will be able to do my work easily without me too.
I doubt this displaces good designers,
I would imagine horse breeders and sailing ship manufacturers thought the same. LLM companies have relased that there's no money in "doing the drudgery work" - like they promised, but only in the 'creative' arts. This is our future now, LLM are already creatiing everything from film, books to art and TV. As @cougar points out, this is the worst it'll ever be, and they stole human's work to do it. Even my wife's published academic works has been used, she's currently part of the Anthropic settlement. They've probably stolen @Binners works as well.
The issue I have is there’s no Intelligence in it , you’ll get a huge load of tat generated when you could do something more elegant and simpler in HTMX but AI doesn’t innovate ,it just regurgitates.
Isn't that what programmers have been doing for years anyway? 😁
One area that AI is a considered a no go is social media influencer/creator/UGC campaigns most work will be rejected or looked down upon, even told not to use filters. I only use AI to help with writing stuff as sometimes it just cleans up the crap I might have written. I would never use it for brands though as above it's too lazy.
TBH you get whatever’s the latest and greatest on medium 🙂
Regardless of whether it’s suitable for you size/application.
@chambord Haven't you just explained why your job isn't redundant though; the skill of structuring the prompt that produces a good output is going to be in demand isn't it?
Kind of agree for now, but the overall impact will be a contraction of the jobs market, less jobs for younger/junior engineers. Eventually I suspect I will no longer be required because a more senior/experienced engineer will be able to do my work easily without me too.
It’s an odd one and depends on the whims of your management , it could just be a race to the bottom to just cut costs, crap code that just about works but looked after by cheaply available and relatively unskilled labour.
When they do get a real problem then they call in an AI Whisper to sort it out and then off they go again.
I think we're starting to see a bit of blow back in regards to AI use - in that in an employment context AI has no responsibility for its mistakes - if you rely on, sell, publish, implement AI output the oversights and shortcomings are all yours.
Ask the chief of West Midlands Police for instance, or lawyers who are putting forward completely fictional case law forward in court. Or the mural artist in Glasgow who's rendering for a Scottish landscape featured a Bald Eagle.
AI will make all the same errors that lazy people make, and given that lazy people are attracted to AI they're the people who'll fail in their diligence to verify the outputs they're given
Alll the various AI / LLM services are vying for dominance and becuase of that they all have an inbuilt seduction- they're configured to flatter the user with their output. Users feel too good about the outputs they receive to properly interrogate them. But the egg ends up on their face if they put that output infront of any more discearning audience.