AI noob here. I saw a YouTube advert for an AI platform in which you pop in a few lines and it spits out an image. I gave the free trial a go and was so impressed on how quickly I could produce something half decent with the AI features + text tools that I paid for a monthly plan. Unfortunately that limits you to 30 AI credits per day while the Pro plan is 100 credits. Before I upgrade yet again, I'm wondering how unique the output would actually be... presumably fed by a huge quantity of data. What its the likely hood that without a very niche request it'll just produce the exact same image again for someone else?
Try logging in with different accounts and using the same text to generate the image? Maybe on a different IP address.
Exact? Unlikely. Very similar? Highly likely. There's also a decent chance it'll hand you something that breaches someone's copyright somewhere. Even better, the output has been agreed to be, in the US and UK at least, uncopyrightable. There's nothing to stop someone taking an image you told it to generate and using it as their own. In short: under no circumstances should you attempt to use a gen-AI system for commercial purposes. See the Demote thread for more on the topic. EDIT: Back at my desk now, so I can provide better references than "go look in Demote": Gen-AI will invent things that don't exist, which is why you'll find people with two heads or 12 fingers. If you're lucky, you'll notice that before you use the image commercially; if not, whoops. The images they generate are only convincing at first glance... and will in no way represent anything from the real world in any kind of convincing manner. But here's the killer: image generation engines will happily spit out images which will get you in serious trouble with copyright holders, even if you don't ask them to. Until it got blocked - and they banned the researcher, twice, with no refund - Midjourney would respond to "screencap" with replicas of randomly-chosen Hollywood blockbuster frames - Marvel, Spider-Man, Frozen. Games, too. If Spider-Man pops up, I'm sure you'll notice - but how confident are you that you could recognise any frame from any Hollywood film before Disney's legal team comes a-knockin'? Not that confident? Yeah, don't use gen-AI for anything public, then.
Interesting… Thankfully not looking to do something as mega, most of the time I’m just playing around. But I have been playing with image generation and the variety of text and image tools to create stuff. Maybe some T Shirt designs or a few art pieces but nothing (hopefully) to put me in court with the lawyers of marvel. Just wouldn’t want to walk into a primark and find the same imagine!
Please don't do this. The market is already flooded with terrible gen-AI tat. Don't add to it. You cannot get "art" by typing a prompt into a Plagiarism-as-a-Service engine. Gen-AI will not make you rich, 'cos a load of grifters have beaten you to it and are doing it at a much bigger scale. I'm rooting for their downfall; don't make me root for yours too.
More pragmatically, I think if it was something for personal use i.e. you’re playing around with image gen looking for something unique that you like and think ‘hmm, that’s cool I’d like to hang that on the wall or print it on a t-shirt’ you’re probably fine. But, as I understand it, it’s not just a case that the same phrase might throw up the same image for someone else, a different phrase could potentially throw up the same image based on the algorithm and training base of the image processor. Though recognise this is a very low probability. Though totally with Gareth on not using it for anything commercial. I think there is too much risk there of either inadvertently using copyright material, which you’ll only found out about too late, and/or ripping off an actual artist somewhere who’s work has been fed into the training without their knowledge or consent. But just my twopenne’worth.
