AI movie posters | Hacker News Reader

AI movie posters

noahveltman.com

278 points_yoqn112 comments

It was less fun for me because I haven't seen many of those movies. If it were 100% classics it would be more fun IMO. Not classic as in old, just most popular. On the other hand who gives a shit what I think. This person probably just generated the images for their favorite movies.

For the ones I did know, I almost always could see a similarity to the "feel" of the movie. But the Rear Window one?? What happened there? Didn't capture the "feel" whatsoever IMO. Of course there are windows but it looks more like a backalley in Japan. The color palette is way off.

Tangentially related, this AI image generation account on Twitter never ceases to fascinate me. Many of the photos are seeded with an initial image however, which reduces the magic/specialness for me a bit: https://twitter.com/images_ai

The color palette is definitely off, but it does capture the sense of looking out a window into other windows. Given that it was primed with a description, it's not surprising that it couldn't get the right colors.

Agreed, I only knew a handful of these films.

Nonetheless, this was super fun.

With many pictures I instantly got the feeling of what the movie is without really realising why. Could this be because the AI is probably trained with human made pictures from ex. flickr (in most cases after the movie), so the AI will generate pictures that basically resemble how people see/remember the movies?

For example the very strong color scheme in Miami Vice, which is basically 100% something that emerged in popular culture after the movie.

So when the AI is generating the pictures based on the plot, could it be deep down generating the pictures partly based on how we people remember the movie?

I had the opposite experience, I couldn't identify any. Maybe I'm going senile.

Same here. After looking at the movie name, I can see the connection, but it's just not strong enough for me to identify the movie from the picture.

Not that this should matter in this context, and no idea how common or uncommon this particular "skill" is, but I can often identify a movie that I've seen years, even decades ago, by seeing just a few seconds of it. I guess this is a form of photographic memory. But, it's not only the raw snapshots, instead it's a mix of image quality, content and if I catch a glimpse of it, the story and characters too. Is there such a thing is "moving pictures memory"?

Edit: I did identify "Being John Malkovich (1999)"! Then again, his face appears in the picture.

> no idea how common or uncommon this particular "skill" is

No idea either. But add one more if you’re counting, it happens to me as you describe.

Or used to. Television was an important trigger because it shows movies unprompted; switching to alternatives where I have to consciously pick what to watch removed the opportunity for the effect to manifest.

I could see Ocean's Eleven and Wizard of Oz, but anyway the other movies felt a bit obscure so nothing else.

The last one, Star Wars, I think needs some deep inspection to get it and is missing the Vader ( which I think anyone with identify with the movie)

It's so interesting how we all seem to recognize these images in different ways. I found a lot of these hard to identify, but Star Wars was an easy one. It stood out because it looked like star troopers and the bottom right looked like Leia.

The ones that I could identify: * Star Wars * Edward Scissorhands * John Malkovich * National Treasure * Hot Tub Time Machine * Willy Wonka * Space Jam * The Hurt Locker * Midsommar (this one gives me flashbacks, btw) * Mad Max * Fear and Loathing * Oceans 11

Everything else was too obscure. Curious which ones other people could identify.

It is a trippy experience because many objects don't make sense, it's like solving a puzzle without having the reference image. I haven't watched about half of those movies, but of those I could instantly recognize Mad Max, Bill & Ted, Black Swan and Star Wars. The rest were enigmas. Maybe bigger images would make things easier.

I think many of the original posters (or stills from the movie) were just in the training data, to be honest.

Willy Wonka & the Chocolate Factory in particular seems to be a direct reconstruction of features from that movie.

To me the wizard of Oz was very obvious and I've never seen the movie, just children's books and maybe a few screenshots. But it's still strange; for example ocean's eleven was immediately my first thought for the according image, but I don't even know why. I don't recognize anything specific in that image, and it doesn't have a distinct set of colors either.

People in tuxedos, with guns and piles of cash.

"With many pictures I instantly got the feeling of what the movie is without really realising why"

I immediately got a feeling of what the movie is about, but a lot of times I named the wrong movie or couldn't remember it. The feeling was similar to when you have something at the tip of the tongue..

I missed most of them (the majority I'd never even seen the movie), some were more obvious (as you mention Miami Vice), but the Fear and Loathing one was very strange to me. I got that one near instantly after seeing but, but unlike the others, was unable to see exactly why

I have seen maybe half of these films. I was able to correctly name some of them immediately upon looking at them: Wizard of Oz, The Matrix, The Big Lebowski.

Some were obvious after reading the title: Star Wars, Harold & Kumar Go to White Castle (never saw it), Being John Malkovich, Space Jam (never saw it), Miami Vice (never saw it), Office Space.

Some that I half guessed or felt like I knew the name of the film but could not immediately recall the name: Castaway, Mad Max: Fury Road, Edward Scissorhands (Thought it was The Nightmare Before Christmas but Tim Burton so close enough)

Obvious ones: Willy Wonka & the Chocolate Factory

Yeah, looking at the Edward Scissorhands one, it was obvious that it was a Tim Burton movie, but not which one. I had guessed Corpse Bride.

Willy Wonka and the chocolate factory looked almost as good as a blurred picture from the movie, especially Willy Wonka’s face

On the other hand, the green and black color scheme that The Matrix made super famous is entirely missing from that one.

Yeah tough the same after I guessed "The Prestige" when it doesn't seem I should have been able to.

This was exactly my experience. On many, my first thought was the correct movie. Very interesting.

This is neat! I was fairly surprised when a few that were instantly in my mind were actually correct - Back to the Future, and Wizard of Oz, Ghost Busters, Big Lebowski, to name a few.

And I don't think its simply because of the likelihood of guessing a popular movie, as many of the titles I've never heard of. Although some are a giveaway by their content and focusing on an obvious subject - Ghost Busters has clear ghosts; Hurt Locker has a bunch of bomb-disposal guys. Others like Edward Sissorhands, have no giveaways but just embody the style; I'd love to see the descriptions provided to the AI

> I'd love to see the descriptions provided to the AI

Unless the descriptions were particularly detailed, I would expect that a lot of this comes from the training data and the descriptions are just prompts for the model to recall which film it is.

For instance, Willy Wonka & the Chocolate Factory (1971) is clearly based on this real poster:

https://www.amazon.com/-/es/Póster-Online-Willy-Chocolate-Fa...

This seems less “generate a poster from a description” and more “recall half-memorised posters“.

Due to how VQGAN + CLIP works, it can't memorize its inputs in the way language models like GPT-3 do.

VQGAN does the generation work, CLIP just says if it's good not, improve the latents, repeat. Here's a good technical writeup: https://ljvmiranda921.github.io/notebook/2021/08/08/clip-vqg...

And of course, the most-common VQGAN was trained on ImageNet, which likely doesn't have every movie poster as training data. (it could be in CLIP though)

What do you suppose the mechanism is for the Charlie and the Chocolate Factory image having a golden ticket held aloft by somebody’s right hand, with a person in a purple outfit and top hat? The page says:

> a brief text description of a movie

However apart from the existence of a golden ticket, I wouldn’t expect those details to make it into a brief description of the film. And yet there’s an original poster matching those details that the VQGAN + CLIP generated image seems to draw from.

Even more convincing to me is the face of John Malkovich being on the poster of Being John Malkovich. Unless the description includes a pretty accurate description of his face (hairstyle, gender, age, facial hair, skin color), the model must have encountered his appearance in its training set.

>(hairstyle, gender, age, facial hair, skin color)

That's not enough for reconstructing the face of John Malkovich from text, you need minute facial feature parameters (eye shape, nose shape, eye-nose distances etc etc)

Because he is famous on the Internet, CLIP “knows” what John Malkovich looks like. Or, more accurately: what an image people would label “John Malkovich” feels like.

Wouldn't the most obvious explanation be a description which mentions Willy Wonka's chocolate factory, which doesn't really turn up anywhere in the training data except the original film media?

Star Wars is an interesting example because it appears to include elements lofted directly from the film (bits of stormtroopers body) alongside a princess who definitely isn't Leia. The algorithm might be creating things from scratch at a high level, but the constituent elements are pretty clearly close reproductions of parts of the source material

Would be interesting to see how well the newer CLIP guided diffusion model works. This is a collection of what it generates with the prompt 'mad max alien spacecraft landed in the desert'

https://i.imgur.com/A1sAaev.jpg

That makes much more sense, pretty typical usage of ML as a blunt instrument.

The interesting thing to do, I think, would be to have data set of general images, and use a movie description to pull from those images.

> Others like Edward Sissorhands, have no giveaways but just embody the style

When I looked at that one, I actually "saw" a figure holding distorted scissors imagery.

Trippy stuff!

Me too! I just knew right away.

The same for me - I got 6 correct with some thought, but three of them were instant ideas (less than a second), and those ones were always right: Cast Away, Space Jam, and Monty Python and the Holy Grail.

Wizard of Oz was one I instantly got but didn't know why. I'm not sure the reason...

The matrix I answered as John Wick, which is pretty interesting.

Fear and Loathing was the other I got but I only tried 10.

The Wizard of Oz is very recognizable by the witch, the tornado and the house flying in the tornado.

For me those are not objects upon individual examination except the house - the witch is barely identifiable as a witch; its generally a triangle body with a hat not at all shaped like a witch's typical pointy one. There are cloud forms, but I'd be hard pressed to call them a tornado. Only when I take in the whole image in-general do I see them, but to focus on specific areas, definitely not.

The way people's brains are analyzing and picking apart these images differently is fascinating to me

Yeah after looking at it closer. I was amused as I thought Wizard of Oz before I realize what the shapes were.

It’s the only one I got. :(

I instantly thought "Matrix" when I saw that one and I suspect it's because I never saw John Wick or many films like that since.

I picked "The Terminator" for the Matrix one, which I thought was thematically close.

I'm pretty impressed and amused by this.

My high point was correctly identifying 'Call me by Your Name.'

Thought process: "Making out on a beach? Oh that looks like an Italian flag!"

To me they are too glowy and they have a weird thing that looks like a mix of chromatic aberration and color bleed when the registration is off.

I thought the Back to the future one was Bladerunner... :)

Pretty cool but no dice for me. First I was intrigued but then they all looked the same. I saw a lot of these movies and tried to relate to the poster, but I could not. Then I thought , would I watch the movie if this were the poster on Netflix or Prime or at a theater and I knew I would not. It feels very much like abstract art, which I don't particularly enjoy.

I also missed the personal connection and the prompt I get from seeing human faces, actors I like and enjoy.

I’ve seen most of the movies and could only guess a few, usually by identifying some specific image in the poster and then deducing the movie, like the ghosts and library for Ghostbusters. Some did a better job of getting the look and feel down, for example I said Forest Gump for Field of Dreams, which are pretty similar.

> but then they all looked the same

Yes, I think I detect a definate bias towards a Picasso/cartoonesque style :-)

Even Dr Stangelove is in colour.

I think our jobs are safe for the forseeable future.

Øyvind Thorsby also published a similar (somewhat easier) quiz a couple weeks ago:

https://thorsbyprojects.tumblr.com/post/660029925870518272/a...

> I typed in movie titles and below is the art I got.

It's clearly impossible to end up with a white haired man and a Delorian from the words "Back to the Future", so it likely has way, way more information about the movie than the title. Possibly even movie posters. I think it's a bit misleading.

The training set could have included images from or related to the movie? It's a cult classic, so it does not seem entirely unlikely to me.

There are signs of memorization of images (or text describing them) from the actual movies. Like in Wall-E and Matrix, you can see, more or less, the robot shape and the green-digits wall in Matrix.

Right. Moreover, it’s visibly memorizing specific images from the movies themselves. It’s very evident for Midsommar and Mad Max, where the renderings contain very idiosyncratic images from these movies which I can’t imagine emerging from mere descriptions of them. So I can only assume that the synopses for these movies contained the actual movie names.

and definitely Space Jam!

This would be so much better if we knew the description fed into the "AI".

My best (frankly uneducated) guess would be that they trained the model with famous movie posters as the objective, and still frames of the movie as the training data. Then they gave frames of these movies to the ai to get a poster out.

I don't think they look as if movie posters were the training objective. No texts, no large faces of leading actors, wrong aspect ratio, unusual colour palette.

Oh, so image data? I got the impression they used a text-only description as input.

I have no idea how, but I actually guessed The Prestige... for some reason I just thought it reminded me of a Tesla Coil. Field of Dreams was a guess just based on it looking like a baseball field.

I'm pretty impressed by ones like Ferris Bueller's Day Off having what I'm pretty sure is the Cubs logo and Wrigley Field, plus Bill and Ted's Excellent Adventure having multiple phonebooths.

To my eye it looks like the Bill and Ted one is trying to include multiple blue police boxes. I'm assuming the movie description included "time travelling phone box" or something like that and the AI generated a bunch of Tardises.

So... this is clearly Art (with a capital A). If I'd been told this was the work of a new artist, I'd believe it, and I'd think they were talented.

What does that mean for the philosophy of Art? Art is meant to communicate emotion [0] but how can something that has no emotion communicate it? And if we exclude this from being Art because of that, then how do we tell "real" Art from this, when there's nothing to distinguish between them?

What happens to Art when you can get an off-the-shelf ML algorithm to produce original, creative, works like this on demand? And what happens to commercial artists and illustrators once we train the algorithm to do their job?

Though that's not all bad - Art could become a hobby done for the joy of creating, rather than a profession done to make money.

Interesting times.

[0] https://en.wikipedia.org/wiki/Art_and_emotion

> And what happens to commercial artists and illustrators once we train the algorithm to do their job?

Perhaps the same as what happened to artisans such as some makers of furniture, shoes, and bespoke cloting. They charge a premium for the "authenticity" of their products. Artisanal, single-source, small batch art.

Yeah true. Obvious question, really. And I guess we'll all face this question sooner or later...

There's human input all over thus, from the algorithm creation to the curating, so I'm not sure this spells the end of art any more than Jackson Pollock using gravity and serendipity to determine where the paint ended up or Renaissance fresco artists having entire studios to generate "their" art. Or the invention of the camera.

Whilst it's aesthetically pleasing in a semi abstract kind of way and you can definitely imagine people paying to put it on their wall, nobody's going to confuse it for the actual human created film posters, so I think the people that make film posters for a living will be just fine (unless all future film directors and promoters want their film posters to be a collage of objects from the film on a psychedelic abstract background, I guess)

I think it's rather interesting how this tech will change the work of human artists. By itself, AI art remains somehow limited in its development.

Decorative art creators might be threatened. But they'e been threatened by reprints for a long time. Political/philosophical art not so much as they are more cognitive and require the artist's societal involvement in some way (even as outcasts).

These AIs generate a specific style of art and you can put it up on your wall as you can put up a reprint of human art. But when they're used with purpose, they can become tools of human artists.

Is there any actual limit in AI-generated Art though?

I'm no expert, but I couldn't tell this generated art from an actual artist (e.g.) "experimenting with abstract representations of movie themes". Could an expert tell them apart? Are there any obvious "tells" for AI-generated art, and can that be overcome with better training?

I feel like we're watching the equivalent of the first computer chess games, with people saying "computers will never beat a grand master because they don't understand the strategic complexity".

People still play chess, despite the fact that their phone can beat them every time. Maybe it doesn't matter that a computer is a better artist, either.

I have been experimenting with this system for months and have concluded that it is a very powerful paint brush. Powerful like steering an elephant. Not like being Superman.

The system is garbage in garbage out. So, feeding it random noise inputs will give noise results. The human factor is “prompt engineering” the inputs to try to coax the elephant in the desired direction. And, in evaluating the outputs.

Sometimes it feels like asking a child to draw something for you. I spent some time trying to get “storm clouds made of lava” but the child insisted “No, dad! Lava stays on the ground. Like THIS!” https://www.reddit.com/r/bigsleep/comments/o033x1/playing_wi...

Similarly, while trying to make a “tempest in a teapot”, instead of a storm, the child was a fan of the My Little Pony character Tempest Shadow. https://www.reddit.com/r/bigsleep/comments/o04sin/trying_to_...

That's awesome :) Thanks for the info. I love the images. And the metaphor of it being a powerful tool for creative expression rather than an actual creative expression itself.

Do you think this will change and get better? Would it help if the AI was capable of interactively refining the inputs (I guess by asking questions instead of making assumptions)?

I think that the tool will get higher quality results with easier to control inputs. Some of that is already happening rapidly.

In theory, given enough machine power, I could see a branching, interactive interface. Like if there are multiple major hills in the latent space ("Tempest-> a storm cloud" vs. "Tempest -> a cartoon pony") the system could identify them all and generate a set of images with different biases towards each hill. Then the user can pick which direction to climb.

The current system on high-end hardware takes about a minute before results for one image even begin to be recognizable. And, about 3 minutes before you can be somewhat confident how it's generally going to turn out.

You won't be able to tell the difference when you go picture frame shopping at IKEA, sure (you could probably even have an on-demand printer which generates art suited to your needs). But you won't be able to invite the AI for an interview about the meaning of its art either. I think someone feeding an AI with movie titles and arranging it in an interesting or meaningful way becomes an artist with the AI as their work material.

Just like chess AIs are the result of human work by the way. I don't know if there is a league of AI chess, but if there were, you'd always have the developer teams behind them who make strategic decisions on how to develop the AI. They become chess players of another kind but chess players nonetheless.

These are very cool. I didn't get any of them, but paired with the titles, many could be art. I'd like to understand what the training set was, and what prompts were given. Some, like space jam and willy Wonka, seem like they must be accessing images from those movies that were in the training set.

It's cool as something generated by an AI, but quite poor in terms of art. I don't like them, and they are boring.

Some of these are just bizarre in how effective they are.

I have never picked up the book or watched the movie Fear and Loathing in Las Vegas. I probably couldn't have told you anything about them beyond what was included in the text description the AI got. Yet I instantly knew that was the movie.

I could even envision a few of these as art work for a special edition blu-ray or something.

In the poster for Bill & Ted's Excellent Adventure, it looks like the AI caught "time travel" and "phone booth" and tried to draw a whole bunch of TARDISes.

Back to the Future has what looks like a TARDIS as well. Although… so does The Matrix, for some reason. Phone booths did play a minor role in that movie, but I kind of doubt that’s the reason. It would be interesting to see the exact text given as the prompt for each movie.

IMO it looks like a booth with ads for iphones on them. It just has no idea what a phone booth is.

https://www.youtube.com/watch?v=lPu2YRywmsI This is the roughly the same idea but applied to the Lyrics of Holland, 1945 by Neutral Milk Hotel.

It's very fascinating to watch and almost empathize with the computer as to how it joined these concepts together (i.e. "bullets fly" ends up like a kind of bullet-hummingbird type creature)

It would be interesting if the AI would be fed the whole movie and create a poster based upon the scenery, characters and locations.

I observed that kind of experiments for a long time, it is catchy, relatively easy to implement and it gives that feeling that artificial intelligence create something by itself. But as long as it will stricly dependent of distribution of the data that we put in it, It will always be some kind of less or more visible modification of input data.

I had a ton of fun with this one!

There is something sort of magic about these images. The satisfaction I felt when I saw a picture and instantly had a thought that turned out to be correct was really refreshing.

It's not the same feeling you get when you put a lot of effort into solving a problem and eventually getting there, but more of an artistic sense. Interesting.

I'm not sure why I got the ones I did, but I got Cast Away, Space Jam, Willy Wonka, National Treasure, Being John Malkovich, and Star Wars. Not sure if they got easier toward the bottom or if I started catching on after viewing the answers as I went down the page.

These images make me wonder what the "brief text description" is for each.

I definitely noticed that it got easier for me after 3-4 tries. I got a sense for how the computer was depicting stuff.

I got nine of them right.

SPOILERS BELOW

I got most of them by looking at either landscapes or color palettes. In some cases, certain characters were actually visible in the collage.

Wizard of Oz, Ghost Busters, Fear and Loathing in Las Vegas, Mad Max: Fury Road, Office Space, Space Jam, Willy Wonka and the Chocolate Factory, National Treasure, Star Wars.

I guessed fury road, fear and loathing, national treasure, Malkovich Malkovich and a handful of others. For some the posters resemble scenes from the movie, others seem to more literally express the title. To me these scream AI generated they remind me a lot of the gan generated visual dreams...

That was fun! I got 16 right, if I’m allowed to count the wrong Willy Wonka and Mad Max. Many of the movies I didn’t get I’ve not seen and only barely heard of, so I certainly relied on guessing on famous movies where the themes in the picture are prominent.

While I had no success in guessing the movie from the image, I had more success in themes - choosing the images I found the most attractive and discovering to my surprise that they strongly tended to be for movies I liked the most.

An image summary of a movie maybe, not a movie poster though. Almost all posters feature the faces of the characters very prominently (presumably in order of screen time) that seems to have been skipped completely

Honestly, I would like to see movie posters made by 5-year-olds that were given a short description of a movie. It wouldn't be the same as this, but it would be interesting in a similar way.

This was really cool. I was able to guess I think seven of them correctly. Some I didn’t even know why I thought of the movie, I just had the impression and it turned out to be right.

I love the art style. I think it works especially well for dystopias

The ambiguity between the bag of pills and typewriter against a backdrop of the strip is pretty uncanny. I would like to use this tool and models for blog post hero image art.

Well. I guessed just one of them - Mad Max. However I've seen probably not more than 10 of those movies, so it's 10% success rate. Not so bad ))

It was surprisingly easy to see the connection to the movie after the fact. But I could guess none of the titles just from the poster.

I would like to see the actual "brief text description of a movie" given to the AI for each image, too.

Spoiler alert:

var answers = document.querySelectorAll('.button'); answers.forEach( answer => { answer.click(); });

I'd say I got ~%20 of them right. Some really interesting images out of them. The one for mad max is beautiful.

The one for Fear and Loathing in Las Vegas could actually be a cover image for that movie. It's perfect.

What were the prompts for each poster?

The Point Break one is fantastic.

The Mad Max one is pretty close.

Indeed. The only one I've got right

Nope! That's a fever dream.

An alternative version of the page should exist that says Fear and Loathing in Las Vegas for every single image.

And all dreamed up by the same person.

Impressed by the prominence of the peaches in the Call By My Name poster.

This was kind of fun for me. Guessed 12 of the movies.

The show answer button doesn't work on Brave mobile

Lodge a bug.

So obvious. I got all the answers correct. Mind blown.

You're a bot, ain't you?

No :)

Got one. I guess it’s clear, we don’t need any more movie poster designers.

I would have appreciated some more recent ones though. I don’t quite remember all movies from the 80s.

I got 5 right. Harold and Kumar, Space Jam, Cast Away, Willy Wonka, and Fear and Loathing. I feel like the prompts were generated off of movie frames? Space Jam for example was very obviously containing cartoons wearing Space Jam jerseys

Reply on news.ycombinator.com