Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I recreated famous album covers with DALL-E

lucytalksdata.com

Post reply on HN
Print view View original

121–130 of 137 posts

Re: I recreated famous album covers with DALL-E

#121

Post by bagels »

2OEH8eoCRo0 wrote: post #80
powersnail wrote: post #35

DALL-E is still highly probabilistic in its judgement. For instance, in this article, it keeps putting "fire" in the the background on something that is likely to be on fire, rather than lighting up the person. I have a similar experience. In my own experiment, I can't get DALL-E to turn off the street lamp at a bus stop in the darkness. I've tried "no light", "broken street lamp", etc.; no use. Any mention of "stree…

The person is on fire. On top of the fire in the image.

I agree, the picture of the man is literally on the picture of the fire. Might have said one man is engulfed in flames, is immolated, alight, or other different phrasings.

Re: I recreated famous album covers with DALL-E

#122

Post by shawabawa3 »

joe_the_user wrote: post #120
whirlwin wrote: post #62

Earlier quoted context omitted.

I have experienced violent or harmful settings to be avoided by DALL-E. E.g. setting a person on fire. Same with drowning - seems to be impossible/hard to generate

It didn't get "swimming to a dollar" right either. I think it doesn't "understand" spatial descriptions unless it happens to find an image match the description.

It definitely struggles with relationships between objects, often confusing them instead (e.g. printing the baby on the dollar bill instead of swimming to it in this case)

Re: I recreated famous album covers with DALL-E

#123

Post by vanadium1st »

xorcist wrote: post #78
vanadium1st wrote: post #39

Earlier quoted context omitted.

I am already doing exactly that, and am getting paid for it. I am a logo artist and I sell pre-made logo designs. Before the current AI services I had to come up with visual ideas by myself, like a caveman. Now I use the AI to generate a bunch of sketches and blurry ideas, and then use my graphic design experience to polish them up to a usable level. Here's how it looks. https://imgur.com/a/DKTsKdC I am absolutely su…

Thank you for that perspective. The linked work is clearly the work of a skilled professional. I am intrigued by the use of AI as a form of creativity assist. As someone without any talent for this, the left pictures are useless for me, as I don't know how to take them into something like the pictures on the right. The point of a sketch is to show them to a customer, but if you would show these sketches to me, I woul…

Using existing logos for inspiration (like the NFL ones) is a tightrope because when you get inspired too much you get into legal problems. No one wants a plagiarised or copied logo. Every small borrowed idea, color, composition or a detail is a potential problem. And if you share too much of them - you have an unusable logo. No such problem with AI logos. You can be as much inspired as you want by it, up to a straight copy if you find it good enough.

Re: I recreated famous album covers with DALL-E

#124

Post by Kaibeezy »

cameronh90 wrote: post #89
randymy wrote: post #66

Worth noting that DALL-E automatically “rejects attempts to create the likeness of any public figures, including celebrities". So, you wouldn't be able to get an image that included the 4 Liverpudlians. It does allow you to create fake faces. Might be fun to try and recreate Miles Davis Tutu, Aladdin Sane, Piano Man.

My experience was that if you name a celebrity (and the request isn't blocked) it quite often generated something that has the same general vibe of the target, while also looking entirely unlike them. It reminds me of how TV shows often have a president that resembles the current president in superficial ways, while being distinct enough that they won't get sued. I'd be interested to know why this happens.

There’s a high standard of harm for libel of public figures — knowingly false plus actual malice, iirc. Seems more likely it’s just bad at these details. How would you test it, since it’s not going to have sufficient or reliable source data for ordinary people?

ETA: OK, well, based on the following comments, it has a prohibition on living people, but you can’t libel the dead. So it either is bad at faces or it has a prohibition there too. The article would have mentioned if DALL-E said it wouldn’t render Lennon or Harrison. QED, bad at faces?

Re: I recreated famous album covers with DALL-E

#125

Post by vanadium1st »

SturgeonsLaw wrote: post #112
vanadium1st wrote: post #39

Earlier quoted context omitted.

I am already doing exactly that, and am getting paid for it. I am a logo artist and I sell pre-made logo designs. Before the current AI services I had to come up with visual ideas by myself, like a caveman. Now I use the AI to generate a bunch of sketches and blurry ideas, and then use my graphic design experience to polish them up to a usable level. Here's how it looks. https://imgur.com/a/DKTsKdC I am absolutely su…

Just tested it after looking through your album, and DALL-E seems pretty good for getting a logo concept generated. This is what I got for the prompt "honey badger logo for an NFL sports team" https://imgur.com/a/i2nnii2 Dunno what's going on with that last guy though, perhaps he's had one too many concussions on the field...

Yes, Dall-e 2 is definitely much better at it than other current AI services. But, most people still don't have access to it. Like me for example. I've been in the waitlist from the start, got a cool portfolio - still didn't get in. Maybe I'm just unlucky, or maybe the problem is that I'm in Ukraine. For some reason OpenAi GPT-3 Playground is restricted in our country, so I expect that other OpenAi products might be closed to me for the same reason.

I've seen many examples of Dalle-2 logos similar to yours. It seems like it got at least to a step "good enough to be usable, but with quality on the cheaper side". Which is super impressive and already puts a lot of designers out of work. I sell cheap stock logos on the side, I would feel that loss of income. But right at this moment there's definitely still a quality ceiling for it and the AI didn't put me out of the job completely. If you want a good and expensive-looking logo, you still need a professional human.

I won't be surprised when that changes too and the AI's get to my level. They are already so much closer than I could ever imagine. But at this moment they aren't there.

Re: I recreated famous album covers with DALL-E

#126

Post by waveywaves »

DALL E works really well if you are specific enough. When you don't get the intended result, it helps to identify the element which wasn't generated properly and then improve your description of the same.

"Two men, one of whom is on fire, shaking hands on an industrial lot." can be rewritten as, "Two men, shaking hands, standing on an industrial lot. Person on the right is on fire. Camera is 30 metres away."

You can go into more specifics of the framing and the angle from which you want the picture to be take. By default, DALL E will give you the most realistic generations to your prompts unless you mention "digital art" or a particular art style. I have gotten the best results when generating art instead of photos.

Re: I recreated famous album covers with DALL-E

#127

Post by Kaibeezy »

felipelalli wrote: post #52
andreyk wrote: post #3

I wonder how long the novelty of DALL-E will persist. HN seems to upvote anything titled "I did X with DALL-E". This is a fun post, but it's not that interesting or surprising. Still worth a look don't get me wrong, but personally didn't learn anything new from it. (eg recreating the famous pink Floyd cover with "Outline of prism on a black background in the middle of scene splits a beam of light coming from the left…

I love this kind of bad humor of HN.

Then you’re gonna loooove these computer-generated “jokes”…

http://joking.abdn.ac.uk/jokebook.shtml

Re: I recreated famous album covers with DALL-E

#128

Post by Engineering-MD »

kgeist wrote: post #75
Engineering-MD wrote: post #17

Earlier quoted context omitted.

It’s hard when it was trained on everything pretty much. That’s the same problem as with GPT3. In my mind it’s still brute forcing a solution but instead of endless computation it’s endless examples

Is this "bruteforcing" really different from what our brains do? We see thousands, millions of little things (examples) every day. Then we combine what we've seen into something new. Probably the only difference is that DALLE's training was done once while our brains are trained every day for 80+ years.

I would say yes it is. For humans there is plenty in the world that is completely novel, and we can (and have to) reason from abstractions or first principles. Being exposed to millions of every day things doesn’t give us intrinsic knowledge of writing fiction, algebra, artistic expression etc. instead, we have to apply abstract knowledge and reason. This may well be possible for an AI system, but I haven’t seen GPT3 or DALLE do this.

It’s hard to test this on DALLE or GPT3 as the internet is a summation of all known knowledge in effect. True unbounded Original thought is hard, and given they have seen everything before, it’s impossible to know if it’s original or just seen it previously. It would be interesting in a decades time to see how DALLE or GPT3 deals with novel ideas that it was never exposed to.

Re: I recreated famous album covers with DALL-E

#129

Post by fimdomeio »

There are a lot of articles focusing on how close does DALL-E match some prompt, but I wonder if this is a suboptimal way to explore the medium.

What if you can get a lot more out of it by embracing the unexpected responses. Can it be a tool for exploring lateral thinking? You provide a prompt computer responds with images that are a prompt to human. A baby swiming next to a dolar bill outputs a distorted person face inside a dolar bill with some baby features, could be the start to a rabbit hole of prompts and images where you'll end up with something completly different than your initial expectations.

Re: I recreated famous album covers with DALL-E

#130

Post by PoignardAzur »

avian wrote: post #59
xwdv wrote: post #7

Although AI artists will destroy a lot of jobs, it will also create demand for new jobs for people who specialize in “paint overs” – taking a high concept output created by AI artists and touching it up to perfection. Or perhaps even beyond just a paint over, and into the realm of recreating an entire AI artwork but with a human touch to get details just right. Looking forward to it.

Here's another spin on this: Imagine you're a software developer. In the near future, your manager wants a feature implemented in your company's app. He throws together a short mail with requirements and sends it to the prompt engineering department. The prompt engineers fix a few typos, clean up the grammar and pepper a few secret sauce keywords around like "in the style of firefox", "in the style of kde". This get…

Honestly, kind of yeah.

Something I've come to realise is that I'm much better at fixing other people's code than coding from scratch.

I could definitely work with that kind of tool.

Post reply on HN