[yeknoda]

I've found using these and similar tools that the amount of prompts and iteration required to create my vision (image or video in my mind) is very large and often is not able to create what I had originally wanted. A way to test this is to take a piece of footage or an image which is the ground truth, and test how much prompting and editing it takes to get the same or similar ground truth starting from scratch. It is basically not possible with the current tech and finite amounts of time and iterations.

[jerf]

It just plain isn't possible if you mean a prompt the size of what most people have been using lately, in the couple hundred character range. By sheer information theory, the number of possible interpretations of "a zoom in on a happy dog catching a frisbee" means that you can not match a particular clip out of the set with just that much text. You will need vastly more content; information about the breed, information about the frisbee, information about the background, information about timing, information about framing, information about lighting, and so on and so forth. Right now the AIs can't do that, which is to say, even if you sit there and type a prompt containing all that information, it is going to be forced to ignore most of the result. Under the hood, with the way the text is turned into vector embeddings, it's fairly questionable whether you'd agree that it can even represent such a thing.

This isn't a matter of human-level AI or superhuman-level AI; it's just straight up impossible. If you want the information to match, it has to be provided. If it isn't there, an AI can fill in the gaps with "something" that will make the scene work, but expecting it to fill in the gaps the way you "want" even though you gave it no indication of what that is is expecting literal magic.

Long term, you'll never have a coherent movie produced by stringing together a series of textual snippets because, again, that's just impossible. Some sort of long-form "write me a horror movie staring a precocious 22-year old elf in a far-future Ganymede colony with a message about the importance of friendship" AI that generates a coherent movie of many scenes will have to be doing a lot of some sort of internal communication in an internal language to hold the result together between scenes, because what it takes to hold stuff coherent between scenes is an amount of English text not entirely dissimilar in size from the underlying representation itself. You might as well skip the English middleman and go straight to an embedding not constrained by a human language mapping.

[LASR]

What you are saying is totally correct.

And this applies to language / code outputs as well.

The number of times I’ve had engineers at my company type out 5 sentences and then expect a complete react webapp.

But what I’ve found in practice is using LLMs to generate the prompt with low-effort human input (eg: thumbs up/down, multiple-choice etc) is quite useful. It generates walls of text, but with metaprompting, that’s kind of the point. With this, I’ve definitely been able to get high ROI out of LLMs. I suspect the same would work for vision output.

[kurthr]

I'm not sure, but I think you're saying what I'm thinking.

Stick the video you want to replicate into -o1 and ask for a descriptive prompt to generate a video with the same style and content. Take that prompt and put it into Sora. Iterate with human and o1 generated critical responses.

I suspect you can get close pretty quickly, but I don't know the cost. I'm also suspicious that they might have put in "safeguards" to prevent some high profile/embarrassing rip-offs.

[robotresearcher]

> Long term, you'll never have a coherent movie produced by stringing together a series of textual snippets because, again, that's just impossible.

Why snippets? Submit a whole script the way a writer delivers a movie to a director. The (automated) director/DP/editor could maintain internal visual coherence, while the script drives the story coherence.

[coffeebeqn]

This almost certainly won’t work. Feel free to feed any of the hundreds of existing film scripts and test how coherent the models can be. My guess is not at all

[robotresearcher]

The clips on the Sora site today would have been utterly astonishing ten years ago. Long term progress can be surprising.

[dragonwriter]

> The clips on the Sora site today would have been utterly astonishing ten years ago.

Yeah, and Apollo 11 would have been utterly astonishing a decade before it occurred. And, yet, if you tried to project out from it to what further frontiers manned spaceflight would reach in the following decades, you’d…probably grossly overestimate what actually occurred.

> Long term progress can be surprising.

Sure, it can be surprising for optimists as well as naysayers; as a good rule of thumb, every curve that looks exponential in an early phase ends up being at best logistic.

[hatefulmoron]

In the long run we are all dead. Saying that technology will be better in the future is almost eye-roll worthy. The real task is predicting what future technology will be, and when it will arrive.

Ask anyone with a chronic illness about the future and they'll tell you we're about 5 years off a cure. They've been saying that for decades. Who knows where the future advancements will be.

[bergen]

https://xkcd.com/605/

[sleepybrett]

This will almost certainly be in theaters within 5 years, probably first as a small experimental project (think blair witch).

[runarberg]

The Blair Witch Project was a (surprise) creative masterpiece. It worked with very limited technology to create a very clever plot which was paired with an amazing marketing. The combination of which the world hadn’t seen before. It took some creative geniuses to peace the Blair Witch Project together.

Generative AI will never produce an experience like that. I know never is a long time, but I’m still gonna call it. You simply can’t produce such a fresh idea by gathering a bunch of data and interpolating.

Maybe someday enough AI will be good enough to create shorter or longer videos with some dialog and even a coherent story (though I doubt it), but it won‘t be fresh or creative. And we humans will at best enjoy it for its stupidity or sloppiness. Not for its cleverness or artistry.

[dumbfounder]

Why does the idea need to be generated by AI? Let people generate the ideas, the AI will help execute. I think soon (3-5 years) a determined person with no video skills will be able to put together a compelling movie (maybe a short). And that is massive. AI doesn’t have to do everything. Like all tech, it’s a productivity tool.

[Breza]

I'm suspicious of most claims of AI growth, but I think screenwriting is an area where there's real potential. There are many screenplays out there, many movie plots are very similar to each other, and human raters could help with training. And it's worth noting that the top four highest grossing movies right now are all sequels or film adaptations. It's not a huge leap to imagine an LLM in the future that's been trained on movie writing being able to create a movie script when given the Wicked musical. https://www.imdb.com/chart/boxoffice/

[sleepybrett]

I'm thinking more of a Gibsonian 'Garage Kubrick'. A solitary auteur (or small team) that produces the film alone perhaps without even touching a camera, generating all the footage using AI (in the novel the auteur creates all the footage through photo/found-footage manipulation, or at least thats all we see in text). The script will probably be human written, I'm not talking about an AI producing a film from scratch, rather a film being produced using AI to create all the visuals and audio.

[SamPatt]

It's a tool. The cleverness and artistry comes from the humans, not from the tools they use.

The AI isn't creating the fresh ideas. People are.

[troupo]

You should watch how movies are made sometime. How a script is developed. How changes to it are made. How storyboards are created. How actors are screened for roles. How locations are scouted, booked, and changed. How the gazillion of different departments end up affecting how a movie looks, is produced, made, and in which direction it goes (the wardrobe alone, and its availability and deadlines will have a huge impact on the movie).

What does "EXT. NIGHT" mean in a script? Is it cloudy? Rainy? Well lit? What are camera locations? Is the scene important for the context of the movie? What are characters wearing? What are they looking at?

What do actors actually do? How do they actually behave?

Here are a few examples of script vs. screen.

Here's a well described script of Whiplash. Tell me the one hundred million things happening on screen that are not in the script: https://www.youtube.com/watch?v=kunUvYIJtHM

Or here's Joker interrogation from The Dark Night Rises. Same million different things, including actors (or the director) ignoring instructions in the script: https://www.youtube.com/watch?v=rqQdEh0hUsc

Here's A Few Good Men: https://www.youtube.com/watch?v=6hv7U7XhDdI&list=PLxtbRuSKCC...

and so on

---

Edit. Here's Annie Atkins on visual design in movies, including Grand Budapest Hotel: https://www.youtube.com/watch?v=SzGvEYSzHf4. And here's a small article summarizing some of it: https://www.itsnicethat.com/articles/annie-atkins-grand-buda...

Good luck finding any of these details in any of the scripts. See minute 14:16 where she goes through the script

Edit 2: do watch The Kerning chapter at 22:35 to see what it actually takes to create something :)

[shermantanktop]

I can't upvote this enough. This topic in the media space has generated a huge amount of naive speculation that amounts to "how hard could it be to do <thing i know nothing about>?"

[FranzFerdiNaN]

> "how hard could it be to do <thing i know nothing about>?"

This is most Hacker News comments summarized lmao. It's kinda my favorite thing of this place: just open any thread and you immediately see so many people rushing to say ''well just do X or Y'' or ''actually it's X or Y and not Z like the experts claim''. Love it.

[shermantanktop]

In this case, it’s movies and TV, which most people enjoy. So there’s a superficial accessibility to the problem which encourages this attitude.

Of course, HN being the place that it is, the same type of comments are made about quantum entanglement and solar panel efficiency.

[bunabhucan]

I agree with you.

At the same time I am curious in the "that person has too many fingers" sense at what a system trained on tens of thousands of movies plus scripts plus subtitles plus metadata etc. would generate.

I thought about it for a bit and I would want to watch a computer generated Sharknado 7 or Hallmark Christmas movie.

[robotresearcher]

Of course normally other people contribute to a movie after the writer. My comment mentioned three of the important roles. This whole thread is about tech that automates away those roles. That's the whole point.

[dbspin]

I think you've misunderstood the objection.

Lets pick something concrete. It's a medieval script, it opens with two knights fighting. OK so later in the script we learn their characters, historic counterparts etc. So your LLM can match nefarious villain to some kind of embedding, and doubtless has trained on countless images of a knight.

But the result is not naively going to understand the level of reality the script is going for - how closely to stick to historic parallels, how much to go fantastical with the depiction. The way we light and shoot the fight and how it coheres with the themes of the scene, the way we're supposed to understand the characters in the context of the scene and the overall story, the references the scene may be making to the genre or even specific other films etc.

This is just barely scraping the surface of the beginnings of thinking about mise en scene, blocking, framing etc. You can't skip these parts - and they're just as much of a challenge as temporal coherence, or performance generation or any of the other hard 'technical issues' that these models have shown no capacity to solve. They're decisions that have to be made to make a film coherent at all - not yet good or tasteful or creative or whatever.

Put another way - you'd need AGI to comprehend a script at the level of depth required to do the job of any HOD on any film. Such a thing is doubtless possible, but it's not going to be shortcut naively the way generation an image is - because it requires understanding in context, precisely what LLMs lack.

[robotresearcher]

> but the result is not naively going to understand the level of reality the script is going for…

We can already get detailed style guidance into picture generation. Declaring you want Picasso cubist, Warner brothers cartoon, or hyper realistic works today. So does lighting instructions, color palettes, on and on.

These future models will not be large language models, they will be multi-modal. Large movie models if you like. They will have tons of context about how scenes within movies cohere, just as LLMs do within documents today.

[letmevoteplease]

Shane Carruth (Primer) released interesting scripts for "A Topiary" and "The Modern Ocean" which now have no hope of being filmed. I hope AI can bring them to life someday. If we get tools like ControlNet for video, maybe Carruth could even "direct" them himself.

[spoaceman7777]

This exists already actually. Kling AI 1.5. Saw the demo on twitter two days ago, which shows a photo-to-video transformation on an image of three women standing on a beach, and the video transformation simulates the camera rotating, with the women moving naturally. Just involves a segment-anything style selection of the women, and drawing a basic movement vector.

https://x.com/minchoi/status/1862975323433795726

[Der_Einzige]

Controlnet for video is just controlnet but ran frame by frame resulting in AI Rotoscoping.

[bwfan123]

brilliant take from Ben Affleck on ai in movies..

"movies will be one of the last things to be replaced by ai"

https://www.youtube.com/watch?v=ypURoMU3P3U

including this quote: "being a craftsman is knowing how to work, art is knowing when to stop"

[rossjudson]

It is absolutely true that LLMs do not know when to stop.

[natmaka]

An adequate prompter (human at the prompt) knows when to stop.

[jerf]

That's what I describe at the end, albeit quickly in lingo, where the internal coherence is maintained in internal embeddings that are never related to English at all. A top-level AI could orchestrate component AIs through embedded vectors, but you'll never do it with a human trying to type out descriptions.

[minimaxir]

> Under the hood, with the way the text is turned into vector embeddings, it's fairly questionable whether you'd agree that it can even represent such a thing.

The text encoder may not be able to know complex relationships, but the generative image/video models that are conditioned on said text embeddings absolutely can.

Flux, for example, uses the very old T5 model for text encoding, but image generations from it can (loosely) adhere to all rules and nuances in a multi-paragraph prompt: https://x.com/minimaxir/status/1820512770351411268

[dragonwriter]

> but image generations from it can (loosely) adhere to all rules and nuances in a multi-paragraph prompt

Flux certainly does not consistently do so across an arbitrary collection of multi-paragraph prompts, as anyone whose run more than a few long prompts past it would recongize; also, the tweet is wrong in the other direction, as well, longer language-model-preprocessed prompts for models that use CLIP (like various SD1.5 and SDXL derivatives) are, in fact, a common and useful technique. (You’d kind of think that the fact that generated prompt here is significantly longer than the 256 token window of T5 would be a clue that the 77 token limit of CLIP might not be as big of a constraint as the tweet was selling it as, too.)

[lmm]

> You might as well skip the English middleman and go straight to an embedding not constrained by a human language mapping.

How would you ever tweak or debug it in that case? It doesn't strictly have to be English, but some kind of human-readable representation of the intermediate stages will be vital.

[amelius]

Can't you just give it a photo of a dog, and then say "use this dog in this or that scene"?

[artemisart]

Yes, the idea works and was explored with dreambooth/textual inversion for image diffusion models.

https://dreambooth.github.io/ https://textual-inversion.github.io/

[minimaxir]

Both of those are of course out of date and require significant training instead of just feeding it a single image.

InstantID (https://replicate.com/zsxkib/instant-id) fixes that issue.

[Auracle]

Dreambooth style training is in no way out of date.

If you just want a face, InstandID/Pulid work - but it’s not going to be very varied. Doing actual training means you can get any perspective, lighting, style, expression, etc - and have the whole body be accurate.

[alpha_squared]

How would that even work? A dog has physical features (legs, nose, eyes, ears, etc.) that they use to interact with the world around them (ground, tree, grass, sounds, etc.). And each one of those things has physical structures that compose senses (nervous system, optic nerves, etc.). There are layers upon layers of intricate complexity that took eons to develop and a single photo cannot encapsulate that level of complexity and density of information. Even a 3D scan can't capture that level of information. There is an implicit understanding of the physical world that helps us make sense of images. For example, a dog with all four paws standing on grass is within the bounds of possibility; a dog with six paws, two of which are on it's head, are outside the bounds of possibility. An image generator doesn't understand that obvious delineation and just approximates likelihood.

[int_19h]

A single photo doesn't have to capture all that complexity. It's carried by all those countless dog photos and videos in the training set of the model.

[krainboltgreene]

Actually, it does have to capture all of that complexity because it's a photon-based analysis of reality. You cannot take a photo without doing that.

[fennecbutt]

This is correct and even image generation models aren't really trained for comprehension of image composition yet.

Even the models based off danbooru and E621 still aren't the best at that. And us furries like to tag art in detail.

The best we can really do at the moment is regional prompting, perhaps they need something similar for video.

[echelon]

For those not in this space, Sora is essentially dead on arrival.

Sora performs worse than closed source Kling and Hailuo, but more importantly, it's already trumped by open source too.

Tencent is releasing a fully open source Hunyuan model [1] that is better than all of the SOTA closed source models. Lightricks has their open source LTX model and Genmo is pushing Mochi as open source. Black Forest Labs is working on video too.

Sora will fall into the same pit that Dall-E did. SaaS doesn't work for artists, and open source always trumps closed source models.

Artists want to fine tune their models, add them to ComfyUI workflows, and use ControlNets to precision control the outputs.

Images are now almost 100% Flux and Stable Diffusion, and video will soon be 100% Hunyuan and LTX.

Sora doesn't have much market apart from name recognition at this point. It's just another inflexible closed source model like Runway or Pika. Open source has caught up with state of the art and is pushing past it.

[1] https://github.com/Tencent/HunyuanVideo

[circlefavshape]

Their online version is all in Chinese (or at least some Chinese-looking script I don't understand) ... and they recommend an 80GB GPU to run the thing, which costs ~€15-18k. Yikes, guess I won't be doing this at home anytime soon

[yeknoda]

something like a white paper with a mood board, color scheme, and concept art as the input might work. This could be sent into an LLM "expander" that increases the words and speficity. Then multiple reviews to tap things in the right direction.

[mikepurvis]

I expect this kind of thing is actually how it's going to work longer term, where AI is a copilot to a human artist. The human artist does storyboarding, sketching in backdrops and character poses in keyframes, and then the AI steps in and "paints" the details over top of it, perhaps based on some pre-training about what the characters and settings are so that there's consistency throughout a given work.

The real trick is that the AI needs to be able to participate in iteration cycles, where the human can say "okay this is all mostly good, but I've circled some areas that don't look quite right and described what needs to be different about them." As far as I've played with it, current AIs aren't very good at revisiting their own work— you're basically just tweaking the original inputs and otherwise starting over from scratch each time.

[programd]

We will shortly have much better tweaking tools which work not only on images and video but concepts like what aspects a character should exhibit. See for example the presentation from Shapeshift Labs.

https://www.shapeshift.ink/

[3form]

And I think this realistically is going to be the shape of the tools to come in the foreseeable future.

[echelon]

You should see what people are building with Open Source video models like HunYuan [1] and ComfyUI + Control Nets. It blows Sora out of the water.

Check out the Banodoco Discord community [2]. These are the people pioneering steerable AI video, and it's all being built on top of open source.

[1] https://github.com/Tencent/HunyuanVideo

[2] https://banodoco.ai/

[prmoustache]

The whole point of AI stuff is not to produce exactly what you have in mind, but what you are describing. Same with text, code, images, video...

[szundi]

Sounds like we achieved 50% of AI then. The artifical is there, now we need the intelligence part.

[baq]

Sora should be evaluated on xkcd strips as inputs.

[miltonlost]

The adage "a picture is worth a thousand words" has the nice corollary "A thousand words isn't enough to be precise about an image".

Now expand that to movies and games and you can get why this whole generative-AI bubble is going to pop.

[TeMPOraL]

> Now expand that to movies and games and you can get why this whole generative-AI bubble is going to pop.

What will save it is that, no matter how picky you are as a creator, your audience will never know what exactly was that you dreamed up, so any half-decent approximation will work.

In other words, a corollary to your corollary is, "Fortunately, you don't need them to be, because no one cares about low-order bits".

Or, as we say in Poland, "What the eye doesn't see, the heart doesn't mourn."

[jsheard]

> What will save it is that, no matter how picky you are as a creator, your audience will never know what exactly was that you dreamed up, so any half-decent approximation will work.

Part of the problem is the "half decent approximations" tend towards a clichéd average, the audience won't know that the cool cyberpunk cityscape you generated isn't exactly what you had in mind, but they will know that it looks like every other AI generated cyberpunk cityscape and mentally file your creation in the slop folder.

I think the pursuit of fidelity has made the models less creative over time, they make fewer glaring mistakes like giving people six fingers but their output is ever more homogenized and interchangable.

[samatman]

Empirically, we've passed the point where that's true, for someone not being lazy about it.

https://www.astralcodexten.com/p/how-did-you-do-on-the-ai-ar...

In other words, someone willing to tweak the prompt and press the button enough times to say "yeah, that one, that's really good" is going to have a result which cannot in fact be reliably binned as AI-generated.

[lmm]

I mean, no? None of the AI-generated images managed to be indistinguishable. Some people were much better than others at spotting the differences. He even quotes, at length, an artist giving a detailed breakdown of what's wrong with one of the images he thought was good.

[TeMPOraL]

Did you read the article? Respondents performed barely better than chance. Sure, no one was actually 100% wrong[0]. Just almost always wrong, with a noticeable bias towards liking AI art more.

The detailed breakdown you mention? Maybe it's accurate to that artist's thought process, maybe it's more of a rationalization; either way, it's not a general rule they, or anyone, could apply to any of the other AI images. Most of those in the article don't exhibit those "telltale signs", and the one that does - the Victorian Megaship - was actually made by human artist with no AI in the mix.

EDIT:

Another image that stands out to me is Riverside Cafe. Myself, like apparently a lot of other people, going by articles' comments, assumed it's a human-made one, because we vaguely remembered Vang Gogh painted something like it. He did, it's called Café Terrace at Night - and yet, despite immediately evoking the association, Riverside Cafe was made by AI, and is actually nothing like Café Terrace at Night at any level.

(I find it fascinating how this work looks like a copy of Van Gogh at first glance, for no obvious reason, but nothing alike once you pause to look closer. It's like... they have similar low-frequency spectra or something?)

EDIT2:

Played around with the two images in https://ejectamenta.com/imaging-experiments/fourifier/. There are some similarities in the spectra, I can't put my finger on them exactly. But it's probably not the whole answer. I'll try to do some more detailed experimentation later.

--

[0] - Nor should you expect it - it would mean either a perfect calibration, or be the equivalent of flipping a coin and getting heads 30 times in a row; it's not impossible, but you shouldn't expect to see it unless you're interviewing fewer people than literally the entire population of the planet.

[randomcatuser]

a somewhat counterintuitive argument is this: AI models will make the overall creative landscape more diverse and interesting, ie, less "average"!

Imagine the space of ideas as a circle, with stuff in the middle being more easy to reach (the "cliched average"). Previously, traversing the circle was incredibly hard - we had to use tools like DeviantArt, Instragram, etc to agglomerate the diverse tastes of artists, hoping to find or create the style we're looking for. Creating the same art style is hiring the artist. As a result, on average, what you see is the result of huge amounts of human curation, effort, and branding teams.

Now reduce the effort 1000x, and all of a sudden, it's incredibly easy to reach the edge of the circle (or closer to it). Sure, we might still miss some things at the very outer edge, but it's equivalent to building roads. Motorists appear, people with no time to sit down and spend 10000 hours to learn and master a particular style can simply remix art and create things wildly beyond their manual capabilities. As a result, the amount of content in the infosphere skyrockets, the tastemaking velocity accelerates, and you end up with a more interesting infosphere than you're used to.

[TeMPOraL]

To extend the analogy, imagine the circle as a probability distribution; for simplicity, imagine it's a bivariate normal joint distribution (aka. Gaussian in 3D) + some noise, and you're above it and looking down.

When you're commissioning an artist to make you some art, you're basically sampling from the entire distribution. Stuff in the middle is, as you say, easiest to reach, so that's what you'll most likely get. Generative models let more people do art, meaning there's more sampling happening, so the stuff further from the centre will be visited more often, too.

However, AI tools also make another thing easier: moving and narrowing the sampling area. Much like with a very good human artist, you can find some work that's "out there", and ask for variations of it. However, there are only so many good artists to go around. AI making this process much easier and more accessible means more exploration of the circle's edges will happen. Not just "more like this weird thing", but also combinations of 2, 3, 4, N distinct weird things. So in a way, I feel that AI tools will surface creative art disproportionally more than it'll boost the common case.

Well, except for the fly in the ointment that's the advertising industry (aka. the cancer on modern society). Unfortunately, by far most of the creative output of humanity today is done for advertising purposes, and that goal favors the common, as it maximizes the audience (and is least off-putting). Deluge of AI slop is unavoidable, because slop is how the digital world makes money, and generative AI models make it cheaper than generative protein models that did it so far. Don't blame AI research for that, blame advertising.

[zmgsabst]

A small technical point:

Tastes are almost never normally distributed along a spectrum, but multi-modal. So the more dimensions you explore in, the more you end up with “islands of taste” on the surface of a hyper sphere and nothing like the normal distribution at all. This phenomenon is deeply tied to why “design by committee” (eg, in movies) always makes financial estimates happy but flops with audiences — there is almost no customer for average anything.

I agree with your conclusion.

[etiam]

I like the picture, but I'd be more impressed with the exploration argument if we were collectively actually doing a good job giving recognition to original and substantial works that already exist. It'd be of greater service in that regard to create a high-quality artificial stand-in for that limited-quantity "attention" and "engagement" all the bloodsuckers seem so keen on harvesting.

(And I do blame the advertisers, but frankly anyone handing them new amplifiers, with entirely predictable consequences, is also not blameless.)

[js8]

I read this argument/analogy and the "AI slop will win" idea reminds me of the idea that "fake news will win".

That is based on perception that it is easier than ever to create fake content, but fails to account for the fact that creating real content (for example, simply taking a video) is even much easier. So while there is more fake content, there is also lot more real content, and so manipulation of reality (for example, denying a genocide) is much harder today than ever.

Anyway, "the AI slop will win" is based on a similar misconception, that total creative output will not increase. But like with fake news, it probably will not be the case, and so the actual amount of good art will increase, too.

I think we are OK as long as normal humans prefer to create real news rather than fake news, and create innovative art rather than cliched art.

[robertlagrant]

It's just like when Bootstrap came out. Terrible-looking websites stopped appearing, but so did beautiful websites.

[wongarsu]

And as AI oversaturates the cliched average, creators will have to get further and further away from the average to differentiate themselves. If you pour a lot of work into your creation you want to make it clear that it isn't some cliched AI drivel.

[skydhash]

You will basically have to provide a video showcasing your workflow.

[dragonwriter]

> I think the pursuit of fidelity has made the models less creative over time, they make fewer glaring mistakes like giving people six fingers but their output is ever more homogenized and interchangeable.

That may be true of any one model (though I don’t think it really is, either, I think newer image gen models are individually capable of a much wider array of styles than earlier models), but it is pretty clearly not true of the whole range of available models, even if you look at a single model “family” like “SDXL derivatives”.

[TeMPOraL]

> I think the pursuit of fidelity has made the models less creative over time (...) their output is ever more homogenized and interchangable.

Ironically, we're long past that point with human creators, at least when it comes to movies and games.

Take sci-fi movies, compare modern ones to the ones from the tail end of the 20th century. Year by year, VFX gets more and more detailed (and expensive) - more and better lights, finer details on every material, more stuff moving and emitting lights, etc. But all that effort arguably killed immersion and believability, by making scenes incomprehensible. There's way too much visual noise in action scenes in particular - bullets and lighting bolts zip around, and all that detail just blurs together. Contrast the 20th century productions - textures weren't as refined, but you could at least tell who's shooting who and when.

Or take video games, where all that graphics works makes everything look the same. Especially games that go for realistic style, they're all homogenous these days, and it's all cheap plastic.

(Seriously, what the fuck went wrong here? All that talk, and research, and work into "physically based rendering", yet in the end, all PBR materials end up looking like painted plastic. Raytracing seems to help a bit when it comes to liquids, but it still can't seem to make metals look like metals and not Fischer-Price toys repainted to gray.)

So I guess in this way, more precision just makes the audience give up entirely.

> they will know that it looks like every other AI generated cyberpunk cityscape and mentally file your creation in the slop folder.

The answer here is the same as with human-produced slop: don't. People are good at spotting patterns, so keep adding those low-order bits until it's no longer obvious you're doing the same thing everyone else is.

EDIT: Also, obligatory reminder that generative models don't give you average of training data with some noise mixed up; they sample from learned distribution. Law of large numbers apply, but it just means that to get more creative output, you need to bias the sampling.

[wongarsu]

Video games (the much larger industry of the two, by revenue) seems to be closer to understanding this. AAA games dominate advertising and news cycles, but on any best-seller list AAA games are on par with indie and B games (I think they call them AA now?). For every successful $60M PBR-rendered Unreal 5 title there is an equally successful game with low-fidelity graphics but exceptional art direction, story or gameplay.

Western movie studios may discover the same thing soon, with the number of high-budget productions tanking lately.

[robertlagrant]

I agree. The one shining hope I have is the incredible art and animation style of Fortiche[0]'s Arcane[1] series. Watch that, and then watch any recent (and identikit) Pixar movie, and they are just streets ahead. It's just brilliant.

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

[1] https://en.wikipedia.org/wiki/Arcane_(TV_series)

[naasking]

I was just going to say this. If you have an artistic vision that you simply must create to the minutest detail, then like any artist, you're in for a lot of manual work.

If you are not beholden to a precise vision or maybe just want to create something that sells, these tools will likely be significant productivity multipliers.

[whstl]

Exactly.

So far ChatGPT is not for writing books, but is great for SEO-spam blogposts. It is already killing the content marketing industry.

So far Dall-E is not for making master paintings, but it's great for stock images. It might kill most of the clipart and stock image industry.

So far Udio and other song generators are not able to make symphonies, but it's great for quiet background music. It might kill most of the generic royalty-free-music industry.

[msabalau]

Half decent approximations work a lot better in generating the equivalent of a stock illustrations of a powerpoint slide.

Actual long form art like a movie works because it includes many well informed choices that work together as a whole.

There seems to be a large gap between generating a few seconds of video vaguely like one's notion, and trying to create 90 minutes that are related and meaningful.

Which doesn't mean that you can't build from this starting place build more robust tools. But if you think that this is a large, hard amount of work, it certainly could call into question optisimitic projections from people who don't even seem to notice that there is work need at all.

[Ar-Curunir]

That's just sad, and why people have a derogative stance towards generative AI: "half-decent" approximation removes all personality from the output, leading to a bunch of slop on the internet.

[TeMPOraL]

It does indeed, but then many of those people don't notice they're already consuming half-decent, personality-less slop, because that's what human artists make too, when churning out commercial art for peanuts and on tight deadlines.

It's less obvious because people project personality onto the content they see, because they implicitly assume the artist cared, and had some vision in mind. Cheap shit doesn't look like cheap shit in isolation. Except when you know it's AI-generated, because this removes the artist from the equation, and with it, your assumptions that there's any personality involved.

[whatevertrevor]

I'm not so sure, one of the primary complaints about IP farming slop that major studios have produced recently is a lack of firm creative vision, and clear evidence of design by committee over artist direction.

People can generally see the lack of artistic intent when consuming entertainment.

[TeMPOraL]

That's true. Then again, complaints about "lack of firm creative vision, and clear evidence of design by committee over artist direction" is something I've seen levied against Disney for several years now; importantly, they started before generative AI found its way into major productions.

So, while GenAI tools make it easier to create superficially decent work that lacks creative intent, the studios managed to do it just fine with human intelligence only, suggesting the problem isn't AI, but the studios and their modern management policies.

[hammock]

It’s like how there are two types of movie directors (or creative directors in general), the dictatorial “100 takes until I get it exactly how I envision it” type, and the “I hired you to act, so you bring the character to life for me and what will be will be” type

Right now AI is more the latter, but many people want it to be the former

[troupo]

AI is neither.

A director letting actors "just be" knows exactly what he/she wants, and choses actors accordingly. Just as the directors that want the most minute detail.

Clint Eastwood tries to do at most one take of a scene. David Fincher is infamous for his dozens of takes.

AI is neither Fincher nor Eastwood.

[wcfrobert]

Do artist really have a fully formed vision in their head? I suspect the creative process is much more iterative rather than one-directional.

[skydhash]

No one can have a fully formed vision. But intent, yes. Then you use techniques to materialize it. Word is a poor substitute for that intent, which is why there’s so many sketches in a visual project.

[maxglute]

And why physical execution frequently significantly departs from sketches and concept art. The amount of intent that doesn't get translated is pretty staggering in both physical and digital pipelines in many projects.

[dartos]

Your eye sees just about every frame of a film…

People may not think they care, but obviously they do. That’s why marvel movies do better than DC ones.

People absolutely care about details in their media.

[TeMPOraL]

Fair point, particularly given the example. My conclusion wrt. Marvel vs. DC is that DC productions care much less about details, in exactly the way I find off-putting.

Not all details matter, some do. And, it's better to not show the details at all, than to be inconsistent in them.

Like, idk., don't identify a bomb as a specific type of existing air-fuel ordnance and then act about it as if it was a goddamn tactical nuke. Something along these lines was what made me stop watching Arrow series.

[dartos]

> Not all details matter, some do

This is a key observation, unfortunately generally solving for what details matter is extremely difficult.

I don’t think video generation models help with that problem, since you have even less control of details than you do with film.

At least before post.

[famouswaffles]

The visuals are the absolute bottom of why DC movies have performed worse over the years.

The movies have just had much worse audience and critical reception.

[throwup238]

“A frame is worth a billion rays”

The last production I worked on averaged 16 hours per frame for the final rendering. The amount of information encoded in lighting, models, texture, maps, etc is insane.

[bongodongobob]

What were you working on? It took a month to render 2 seconds of video?

[throwup238]

VFX heavy feature for a Disney subsidiary. Each frame is rendered independently of each other - it’s not like video encoding where each frame depends on the previous one, they all have their own scene assembly that can be sent to a server to parallelize rendering. With enough compute, the entire film can be rendered in a few days. (It’s a little more complicated than that but works to a first order approximation)

I don’t remember how long the final rendering took but it was nearly two months and the final compute budget was 7 or 8 figures. I think we had close to 100k cores running at peak from three different render farms during crunch time, but don’t take my word for it I wasn’t producing the picture.

[dist-epoch]

Are they still using CPUs and not GPUs for rendering?

Weren't the rendering algos ported to CUDA yet?

[jsheard]

GPU renderers exist but they have pretty hard scaling limits, so the highest end productions still use CPU renderers almost exclusively.

The 3D you see in things like commercials is usually done on GPUs though because at their smaller scale it's much faster.

[throwup238]

There's plenty of GPU renderers but they face the same challenge as large language models: GPU memory is much more expensive and limited that CPU memory.

A friend recently told me about a complex scene (I think it was a Marvel or Star Wars flick) where they had so much going on in the scene with smoke, fire, and other special effects that they had to wait for a specialized server with 2TB of RAM to be assembled. They only had one such machine so by the time the rest of the movie was done rendering, that one scene still had a month to go.

[fc417fc802]

I'm not sure how well suited GPUs are to the workload. They're also rather memory constrained. The Moana dataset is from 2016 so it's not exactly cutting edge but good luck loading it into vram.

https://www.disneyanimation.com/data-sets/?drawer=/resources...

https://datasets.disneyanimation.com/moanaislandscene/island...

> When everything is fully instantiated the scene contains more than 15 billion primitives.

[Arelius]

Most VFX productions take over 2 CPU hours a frame for final video, and have for a very long time. It takes shorter then a month since this gets parallelized on large render farms.

[elmigranto]

I would guess there is more than one computer :)

Pixar's stuff famously takes days per frame.

[Arelius]

> Pixar's stuff famously takes days per frame.

Do you have a citation for this? My guess would be much closer to a couple of hours per frame.

[elmigranto]

https://sciencebehindpixar.org/pipeline/rendering

[raincole]

The point is not to be precise. It's to be "good enough".

Trust me, even if you work with human artists, you'll keep saying "it's not quite I initially invisioned, but we don't have budget/time for another revision, so it's good enough for now." all the time.

[beambot]

Corollary: I couldn't create an original visual piece of art to save my life, so prompting is infinitely better than what I could do myself (or am willing to invest time in building skills). The gen-AI bubble isn't going to burst. Pareto always wins.

[Al-Khwarizmi]

If you can build a system that can generate engaging games and movies, from an economic (bubble popping or not popping) point of view it's largely irrelevant whether they conform to fine-grained specifications by a human or not.

[dartos]

In other words:

If you find a silver bullet then everything else is largely irrelevant.

Idk if you noticed but that “if” is carrying an insane amount of weight.

[jsheard]

Text generation is the most mature form of genAI and even that isn't even remotely close to producing infinite engaging stories. Adding the visual aspect to make that story into a movie or the interactive element to turn it into a game is only uphill from there.

[soheil]

Maybe your AI bubble! If you define AI to be something like just another programming language yes you will be sadly disappointed. You see it as an employee with its own intuitions and ways of doing things that you're trying to micromanage.

I have a bad feeling that you'd be a horrible manager if you ever were one.

[GistNoesis]

(2020) https://arxiv.org/abs/2010.11929 : an image is worth 16x16 words transformers for image recognition at scale

(2021) https://arxiv.org/abs/2103.13915 : An Image is Worth 16x16 Words, What is a Video Worth?

(2024) https://arxiv.org/abs/2406.07550 : An Image is Worth 32 Tokens for Reconstruction and Generation

[dartos]

Those are indeed 3 papers.

[GistNoesis]

Yes in a nutshell they explain that you can express a picture or a video with relatively few discrete information.

First paper is the most famous and prompted a lot of research to using text generation tools in the image generation domain : 256 "words" for an image, Second paper is 24 reference image per minutes of video, Third paper is a refinement of the first saying you only need 32 "tokens". I'll let you multiply the numbers.

In kind of the same way as a who's who game, where you can identify any human on earth with ~32bits of information.

The corollary being that contrary to what parent is telling there is no theoretical obstacle to obtaining a video from a textual description.

[dartos]

I think something is getting lost in translation.

These papers, from my quick skim (tho I did read the first one fully years ago,) seem to show that some images and to an extent video can be generated from discrete tokens, but does not show that exact images nor that any image can be.

For instance, what combination of tokens must I put in to get _exactly_ Mona Lisa or starry night? (Tho these might be very well represented in the data set. Maybe a lesser known image would be a better example)

As I understand, OC was saying that they can’t produce what they want with any degree of precision since there’s no way to encode that information in discrete tokens.

[GistNoesis]

If you want to know what tokens you want to obtain _exactly_ Mona Lisa, or any other image, you take the image and put it through your image tokenizer aka encode it, and if you have the sequence of token you can decode it to an image.

VQ-VAE (Vector Quantised-Variational AutoEncoder), (2017) https://arxiv.org/abs/1711.00937

The whole encoding-decoding process is reversible, and you only lose some imperceptible "details", the process can be either trained with a L2Loss, or a perceptual loss depending what you value.

The point being that images which occurs naturally are not really information rich and can be compressed a lot by neural networks of a few GB that have seen billions of pictures. With that strong prior, aka common knowledge, we can indeed paint with words.

[stale2002]

You are half right. Its funny because I use the same same. Mine is "A picture is worth a thousand words. thats why it takes 1000 words to describe the exact image that you want! Much better to just use Image to Image instead".

Thats my full quote on this topic. And I think it stands. Sure, people won't describe a picture. instead, they will take an existing picture or video, and do modifications of it, using AI. That is much much simpler and more useful, if you can file a scene, and then animate it later with AI.

[ben_w]

> Now expand that to movies and games and you can get why this whole generative-AI bubble is going to pop.

The prior sentence does not imply the conclusion.

[meta_x_ai]

A picture is worth a thousand words.

A word is worth a thousand pictures. (E.g Love)

It is abstraction all the way

[gloosx]

it is all Information to be precise.

[8n4vidtmkvmk]

Actually, I've gotten some great results with image2text2image with less than a thousand words. Maybe not enough for a video, but for some not too crazy images, it is enough!

[fooker]

Sure it's going to pop. But when is the important question.

Being too early about this and being wrong are the same.

[szundi]

Comment was probably rather about the 360 degree turning heads etc.

[mrandish]

I agree that people who want any meaningful precision in their visual results will inevitably be disappointed.

[isoprophlex]

And another thing that irks me: none of these video generators get motion right...

Especially anything involving fluid/smoke dynamics, or fast dynamic momements of humans and animals all suffer from the same weird motion artifacts. I can't describe it other than that the fluidity of the movements are completely off.

And as all genai video tools I've used are suffering from the same problem, I wonder if this is somehow inherent to the approach & somehow unsolvable with the current model architectures.

[giantrobot]

I think one of the biggest problems is the models are trained on 2D sequences and don't have any understanding of what they're actually seeing. They see some structure of pixels shift in a frame and learn that some 2D structures should shift in a frame over time. They don't actually understand the images are 2D capture of an event that occurred in four dimensions and the thing that's been imaged is under the influence of unimaged forces.

I saw a Santa dancing video today and the suspension of disbelief was almost instantly dispelled when the cuffs of his jacket moved erratically. The GenAI was trying to get them to sway with arm movements but because it didn't understand why they would sway it just generated a statistical approximation of swaying.

GenAI also definitely doesn't understand 3D structures easily demonstrated by completely incorrect morphological features. Even my dogs understand gravity, if I drop an object they're tracking (food) they know it should hit the ground. They also understand 3D space, if they stand on their back legs they can see over things or get a better perspective.

I've yet to see any GenAI that demonstrates even my dogs' level of understanding the physical world. This leaves their output in the uncanny valley.

[jeroen]

They don't even get basic details right. The ship in the 8th video changes with every camera change and birds appear out of nowhere.

[nonameiguess]

As far as I can tell it's a problem with CGI at all. Whether you're using precise physics models or learned embeddings from watching videos, reproducing certain physical events is computationally very hard, whereas recording them just requires a camera (and of course setting up the physical world to produce what you're filming, or getting very lucky). The behind the scenes from House of the Dragon has a very good discussion of this from the art directors. After a decade and a half of specializing in it, they have yet to find any convincing way to create fire other than to actually create fire and film it. This isn't a limitation of AI and it has nothing to do with intelligence. A human can't convincingly animate fire, either. It seems to me that discussions like this from the optimist side always miss this distinction and it's part of why I think Ben Affleck was absolutely correct that AI can't replace filmmaking. Regardless of the underlying approach, computationally reproducing what the world gives you for free is simply very hard, maybe impossible. The best rendering systems out there come nowhere close to true photorealism over arbitrary scenarios and probably never will.

[soheil]

What's the point of poking holes in new technology and nitpiking like this? Are you blind to the immense breakthroughs made today and yet you focus what irks you about some tiny detail that might go away after a couple of versions?

[Uehreka]

At this phase of the game a lot of people are pretty accustomed to the pace of technological innovation in this space, and I think it’s reasonable for people to have a sense of what will/won’t go away in a few versions. Some of Sora’s issues may just require more training, some of these issues are intrinsic to their approach and will not be solvable with their current method.

To that end, it is actually extremely important to nit-pick this stuff. For those of us using these tools, we need to be able to talk shop about which ones are keeping up, which are work like shit in practice, and which ones work but only in certain situations, and which situations those are.

[benchmarkist]

Neural networks use smooth manifolds as their underlying inductive bias so in theory it should be possible to incorporate smooth kinematic and Hamiltonian constraints but I am certain no one at OpenAI actually understands enough of the theory to figure out how to do that.

[david-gpu]

> I am certain no one at OpenAI actually understands enough of the theory to figure out how to do that

We would love to learn more about the origin of your certainty.

[benchmarkist]

I don't work there so I'm certain there is no one with enough knowledge to make it work with Hamiltonian constraints because the idea is very obvious but they haven't done it because they don't have the wherewithal to do so. In other words, no one at OpenAI understands enough basic physics to incorporate conservation principles into the generative network so that objects with random masses don't appear and disappear on the "video" manifold as it evolves in time.

[david-gpu]

> the idea is very obvious but they haven't done it because they don't have the wherewithal to do so

Fascinating! I wish I had the knowledge and wherewithal to do that and become rich instead of wasting my time on HN.

[benchmarkist]

No one is perfect but you should try to do better and waste less time on HN now that you're aware and can act on that knowledge.

[dartos]

How does your conclusion follow from your statement?

Neural networks are largely black box piles of linear algebra which are massaged to minimize a loss function.

How would you incorporate smooth kinematic motion in such an environment?

The fact that you discount the knowledge of literally every single employee at OpenAI is a big signal that you have no idea what you’re talking about.

I don’t even really like OpenAI and I can see that.

[benchmarkist]

I've seen the quality of OpenAI engineers on Twitter and it's easy enough to extrapolate. Moreoever, neural networks are not black boxes, you're just parroting whatever you've heard on social media. The underlying theory is very simple.

[dartos]

Do not make assumptions about people you do not know in an attempt to discredit them. You seem to be a big fan of that.

I have been working with NLP and neural networks since 2017.

They aren’t just black boxes, they are _largely_ black boxes.

When training an NN, you don’t have great control over what parts of the model does what or how.

Now instead of trying to discredit me, would you mind answering my question? Especially since, as you say, the theory is so simple.

How would you incorporate smooth kinematic motion in such an environment?

[benchmarkist]

Why would I give away the idea for free? How much do you want to pay for the implementation?

[esafak]

There are physicists at OpenAI. You can verify with a quick search. So someone there clearly knows these things.

[benchmarkist]

I'd be embarrassed if I was a physicists and my name was associated with software that had phantom masses appearing and disappearing into the void.

[esafak]

Why don't you write a paper or start a company to show them the right way to do it?

[benchmarkist]

I don't think there is any real value in making videos other than useless entertainment. The real inspired use of computation and AI is to cure cancer, that would be the right way to show the world that this technology is worthwhile and useful. The techniques involved would be the same because one would need to include real physical constraints like conservation of mass and energy instead of figuring out the best way to flash lights on the screen with no regard for any foundational physical principles.

Do you know anyone or any companies working on that?

[beefnugs]

AI isn't trying to sell to you: a precise artist with real vision in your brain. It is selling to managers who want to shit out something in an evening that approximates anything, that writes ads that no one wants to see anyway, that produces surface level examples of how you can pay employees less because "their job is so easy"

[spuz]

Yes and the thing is, even for those tasks, it's incredibly difficult to achieve even the low bar that a typical advertising manager expects. Try it yourself for any real world task and you will see.

[cornel_io]

Counterpoint: our CEO spent 25 minutes shitting out a bunch of AI ads because he was frustrated with the pace of our advertising creative team. They hated the ads that he created, for the reasons you mention, but we tested them anyways and the best performing ones beat all of our "expert" team's best ads by a healthy margin (on all the metrics we care about, from CTR to IPM and downstream stuff like retention and RoAS).

Maybe we're in a honeymoon period where your average user hasn't gotten annoyed by all the slop out there and they will soon, but at least for now, there is real value here. Yes, out of 20 ads maybe only 2 outperform the manually created ones, but if I can create those 20 with a couple hundred bucks in GenAI credits and maybe an hour or two of video editing that process wipes the floor with the competition, which is several thousand dollars per ad, most of which are terrible and end up thrown away, too. With the way the platforms function now, ad creative is quickly becoming a volume-driven "throw it at the wall and see what sticks" game, and AI is great for that.

[sarchertech]

> Maybe we're in a honeymoon period where your average user hasn't gotten annoyed by all the slop out there and they will soon

It’s this. A video ad with a person morphing into a bird that takes off like a rocket with fire coming out of its ass, sure it might perform well because we aren’t saturated with that yet.

You’d probably get a similar result by giving a camera to a 5 year old.

But you also have to ask what that’s doing long term to your brand.

[gonzobonzo]

> Counterpoint: our CEO spent 25 minutes shitting out a bunch of AI ads because he was frustrated with the pace of our advertising creative team. They hated the ads that he created, for the reasons you mention, but we tested them anyways and the best performing ones beat all of our "expert" team's best ads by a healthy margin (on all the metrics we care about, from CTR to IPM and downstream stuff like retention and RoAS).

My guess is that the criticism of AI not being that good is correct, but many people don't realize that most humans also aren't that good, and that it's quite possible that the AI performs better than mediocre humans.

This shouldn't be much of a surprise, we've seen automation replace low skilled labor in a lot of industries. People seem uncomfortable with the possibility that there's actually a lot of low skilled labor in the creative industry that could also be easily replaced.

[mewpmewp2]

A/B/C/D testing is the perfect grounds for that. You can keep automatically generating and iterating quickly while A/B tests are constantly being ran. This data on CTR can later be used to train the model better as well.

[soheil]

You seem to speak from experience of being that manager... I'm not going to ask what you shit out in your evenings.

[minimaxir]

Way back in the days of GPT-2, there was an expectation that you'd need to cherry-pick atleast 10% of your output to get something usable/coherent. GPT-3 and ChatGPT greatly reduced the need to cherry-pick, for better or for worse.

All the generated video startups seem to generate videos with much lower than 10% usable output, without significant human-guided edits. Given the massive amount of compute needed to generate a video relative to hyperoptimized LLMs, the quality issue will handicap gen video for the foreseeable future.

[joe_the_user]

Plus editing text or an image is practical. Video editors typically are used to cut and paste video streams - a video editor can't fix a stream of video that gets motion or anatomy wrong.

[didibus]

Right, but you're thinking as someone who has a vision for the image/video. Think from someone who is needing an image/video and would normally hire a creative person for it, they might be able to get away with AI instead.

The same "prompt" they'd give the creative person they hired... Say, "I want an ad for my burgers that make it look really good, I'm thinking Christmas vibes, it should emphasize our high quality meat, make it cheerful, and remember to hint at our brand where we always have smiling cows."

Now that creative person would go make you that advert. You might check it, give a little feedback for some minor tweaks, and at some point, take what you got.

You can do the same here. The difference right now is that it'll output a lot of junk that a creative person would have never dared show you, so that initial quality filtering is missing. But on the flip side, it costs you a lot less, can generate like 100 of them quickly, and you just pick one that seems good enough.

[hipadev23]

Real artists struggle matching vague descriptions of what is in your head too. This is at least quicker?

[staticman2]

Real artists take comic book scripts and turn them into actual comic books every month. They may not match exactly what the writer had in mind, but they are fit for purpose.

[TeMPOraL]

> They may not match exactly what the writer had in mind, but they are fit for purpose.

That's what GenAI is doing, too. After all, the audience only sees the final product; they never get know what the writer had in mind.

[staticman2]

I haven't used SORA, but none of the GenAI I'm aware of could produce a competent comic book. When a human artist draws a character in a house in panel 1, they'll draw the same house in panel 2, not a procedurally generated different house for each image.

If a 60 year old grizzled detective is introduced in page 1, a human artist will draw the same grizzled detective in page 2, 3 and so on, not procedurally generate a new grizzled detective each time.

[TeMPOraL]

A human artist keeps state :). They keep it between drawing sessions, and more importantly, they keep very detailed state - their imagination or interpretation of what the thing (house, grizzled detective, etc.) is.

Most models people currently use don't keep state between invocations, and whatever interpretation they make from provided context (e.g. reference image, previous frame) is surface level and doesn't translate well to output. This is akin to giving each panel in a comic to a different artist, and also telling them to sketch it out by their gut, without any deep analysis of prior work. It's a big limitation, alright, but researchers and practitioners are actively working to overcome it.

(Same applies to LLMs, too.)

[Der_Einzige]

Btw there’s a way to match characters in a batch in the forge webUI which guarantees that all images in the batch have the same figure in it. Trivial to implement this in all other image generators. This critique is baseless.

[staticman2]

So prove it. If you are in good faith arguing an AI, via automation can draw a comic script with consistent figures, please tell an AI to draw the images in the first 3 pages of this script I pulled from the comic book script archive:

https://www.comicsexperience.com/wp-content/uploads/2018/09/...

Or if you can't do this, explain why the feature you mentioned cannot do this, and what it or good for?

[kevingadd]

I work with professional artists all the time and this is not the case. They're generally quite good at extrapolating from a couple paragraphs into something fantastic, often exactly what I had in mind.

In comparison I've messed around with prompting image generator models quite a bit and it's not possible to get remotely close to the quality level of even rough paid concept work by a professional, and the credits to run these models aren't particularly cheap.

[coffeebeqn]

With real art you can start from somewhere and keep building on that foundation. Say you pick an angle to shoot from and test different actors and scenes from that angle. With AI you’re re-rolling the dice for every iteration. If you’re happy that it looks 80% correct then sure it’s maybe passable.

I think people are getting way ahead of their skis here. Even in 2D I can’t for example generate inventory images for weapons and items for a game yet. Which is an orders of magnitude simpler test case than video. They all are slightly different styles. If I don’t care that they all look different in strange ways then it’s useful - but any consumer will think it looks like crap

[janalsncm]

The point is if you are the artist and have something in your head. It’s the same problem with image editing. I am sure you have experienced this.

[TeMPOraL]

There is no problem unless you insist on reflecting what you had in mind exactly. That needs minute controls, but no matter the medium and tools you use, unless you're doing it in your own quest for artistic perfection, the economic constraints will make you stop short of your idea - there's always a point past which any further refinement will not make a difference to the audience (which doesn't have access to the thing in your head to use as reference), and the costs of continuing will exceed any value (monetary or otherwise) you expect to get from the work.

AI or not, no one but you cares about the lower order bits of your idea.

[throwup238]

Nobody else really cares about the lower order bits of the idea but they do care that those lower order bits are consistent. The simplest example is color grading: most viewers are generally ignorant of artistic choices in color palettes unless it’s noticeable like the Netflix blue tint but a movie where the scenes haven’t been made consistently color graded is obviously jarring and even an expensive production can come off amateur.

GenAI is great at filling in those lower order bits but until stuff like ControlNet gets much better precision and UX, I think genAI will be stuck in the uncanny valley because they’re inconsistent between scenes, frames, etc.

[TeMPOraL]

Yup, 100% agreed on that, and mentioned this caveat elsewhere. As you say - people don't pay attention to details (or lack of it), as long as the details are consistent. Inconsistencies stand out like sore thumbs. Which is why IMO it's best to have less details than to be inconsistent with them.

[pbhjpbhj]

>There is no problem unless you insist on reflecting what you had in mind exactly.

Not disagreeing, just noting: this is not how [most?] people's minds work {I don't think you're holding to that opinion particularly, I'm just reflecting on this point}. We have vague ideas until an implementation is shown, then we examine it and latch on to a detail and decide if it matches our idea or not. For me, if I'm imagining "a superhero planting vegetables in his garden" I've no idea what they're actually wearing, but when an artist or genAI shows me it's a brown coat then I'll say "no something more marvel". Then when ultimately they show me something that matches the idea I had _and_ matches my current conception of the idea I had... then I'll point out the fingernails are too long, when in the idea I hadn't even perceived the person had fingers, never mind too-long fingernails!

I'd warrant any actualised artistic work has some delta with the artists current perception of the work; and a larger delta with their initial perception of it.

[janalsncm]

I disagree. Even without exactness, adding any reasonable constraints is impossible. Ask it to generate a realistic circuit diagram or chess board or any other thing where precision matters. Good luck going back and forth getting it right.

These are situations with relatively simple logical constraints, but an infinite number of valid solutions.

Keep in mind that we are not requiring any particular configuration of circuit diagram, just any diagram that makes sense. There are an infinite number of valid ones.

[TeMPOraL]

That's using the wrong tool for a job :). Asking diffusion models to give you a valid circuit diagram is like asking a painter to paint you pixel-perfect 300DPI image on a regular canvas, using their standard paintbrush. It ain't gonna work.

That doesn't mean it can't work with AI - it's that you may need to add something extra to the generative pipeline, something that can do circuit diagrams, and make the diffusion model supply style and extra noise (er, beautifying elements).

> Keep in mind that we are not requiring any particular configuration of circuit diagram, just any diagram that makes sense. There are an infinite number of valid ones.

On that note. I'm the kind of person that loves to freeze-frame movies to look at markings, labels, and computer screens, and one thing I learned is that humans fail at this task too. Most of the time the problems are big and obvious, ruining my suspension of disbelief, and importantly, they could be trivially solved if the producers grabbed a random STEM-interested intern and asked for advice. Alas, it seems they don't care.

This is just a specific instance of the general problem of "whatever you work with or are interested in, you'll see movies keep getting it wrong". Most of the time, it's somewhat defensible - e.g. most movies get guns wrong, but in way people are used to, and makes the scenes more streamlined and entertaining. But with labels, markings and computer screens, doing it right isn't any more expensive, nor would it make the movie any less entertaining. It seems that the people responsible don't know better or care.

Let's keep that in mind when comparing AI output to the "real deal", as to not set an impossible standards that human productions don't match, and never did.

[janalsncm]

The issue isn’t any particular constraint. The issue is the inability to add any constraints at all.

In particular, internal consistency is one of the important constraints which viewers will immediately notice. If you’re just using sora for 5 second unrelated videos it may be less of an issue but if you want to do anything interesting you’ll need the clips to tie together which requires internal consistency.

[mlboss]

So what I am getting a use-case for brain-computer interface.

[hmottestad]

When I first started learning Photoshop as a teenager I often knew what I wanted my final image to look like, but no matter how hard I tried I could never get the there. It wasn't that it was impossible, it was just that my skills just weren't there yet. I needed a lot more practice before I got good enough to create what I could see in my imagination.

Sora is obviously not Photoshop, but given that you can write basically anything you can think of I reckon it's going to take a long time to get good at expressing your vision in words that a model like Sora will understand.

[corytheboyd]

Free text is just the fundamentally wrong input for precision work like this. Because it is wrong for this doesn’t mean it has NO purpose, it’s still useful and impressive for what it is.

FWIW I too have been quite frustrated iterating with AI to produce a vision that is clear in my head. Past changing the broad strokes, once you start “asking” for specifics, it all goes to shit.

Still, it’s good enough at those broad strokes. If you want your vision to become reality, you either need to learn how to paint (or whatever the medium), or hire a professional, both being tough-but-fair IMO.

[londons_explore]

I don't think it'll be long before GUI tools catch up for editing video.

Things like rearranging things in the scene with drag'n'drop sound implementable (although incredibly GPU heavy)

[ohthehugemanate]

If you have a specific vision, you will have to express the detailed information of that vision into the digital realm somehow. You can use (more) direct tools like premiere if you are fluent enough in their "language". Or you can use natural language to express the vision using AI. Either way you have to get the same amount of information into a digital format.

Also, AI sucks at understanding detail expressed in symbolic communication, because it doesn't understand symbols the way linguistic communication expects the receiver to understand them.

My own experience is that all the AI tools are great for shortcutting the first 70-80% or so. But the last 20% goes up an exponential curve of required detail which is easier and easier to express directly using tooling and my human brain.

Consider the analogy to a contract worker building or painting something for you. If all you have is a vague description, they'll make a good guess and you'll just have to live with that. But the more time you spend with them communicating (through description, mood boards rough sketches etc) the more accurate to your detailed version it will get. But you only REALLY get exactly what you want if you do it yourself, or sit beside them as they work and direct almost every step. And that last option is almost impossible if they can't understand symbolic meaning in language.

[cube2222]

Agreed. It’s still much better than what I could do myself without it, though.

(Talking about visual generative AI in general)

[JKCalhoun]

Yeah, but if I handed you a Maxfield Parrish it would be better than either of us can do — but not what I asked for.

I find generative AI frustrating because I know what I want. To this point I have been trying but then ultimately sitting it out — waiting for the one that really works the way I want.

[cube2222]

For me even if I know what I want, if I’m using gen AI I’m happy to compromise and get good enough (which again, is so much better than I could do otherwise).

If you want higher quality/precision, you’ll likely want to ask a professional, and I don’t expect that to change in the near future.

[adamc]

That limits its value for industries like Hollywood, though, doesn't it? And without that, who exactly is going to pay for this?

[cube2222]

To me, currently, visual generative ai is an evolution and improvement of stock images, and has effectively the same purpose.

People pay for stock images.

[adamc]

Yeah, maybe for some purposes. In business, people sometimes pay for stock images but often don't have the expertise or patience to really spend a lot of time coaching a video into fruition. Maybe for advertising or other contexts where more effort is worth it (not just powerpoints), but it feels like a slim audience.

[jddj]

Advertisers, I guess. Same folks who paid for everything else around here

[adamc]

Yeah, I just question if there are enough customers to make this work.

[joe_the_user]

The thing about Hollywood is that movies aren't made by a producer or director creating a description and an army of actors, tech and etc doing exactly that.

What happens is a description becomes a longer specification or script that's still good and hangs together in itself and then further iterations involving professionals who can't do "exactly what the director wants" but rather do something further that's good and close enough to what the director wants.

[skydhash]

Also, a team of experts and professionals that knows better than the director how a specific thing work.

[diob]

I believe it. I was just using AI to help out with some mandatory end of year writing exercises at work.

Eventually, it starts to muck with the earlier work that it did good on, when I'm just asking it to add onto it.

I was still happy with what I got in the end, but it took trial and error and then a lot of piecemeal coaxing with verification that it didn't do more than I asked along the way.

I can imagine the same for video or images. You have to examine each step post prompt to verify it didn't go back and muck with the already good parts.

[planb]

Iterations are the missing link.

With ChatGPT, you can iteratively improve text (e.g., "make it shorter," "mention xyz"). However, for pictures (and video), this functionality is not yet available. If you could prompt iteratively (e.g., "generate a red car in the sunset," "make it a muscle car," "place it on a hill," "show it from the side so the sun shines through the windshield"), the tools would become exponentially more useful.

[goldfeld]

If you use it in a utilitarian way it'll give you a run for your money, if you use for expression, such as art, learning to embrace some serendipity, it makes good stuff.

[titzer]

As only a cursory user of said tools (but strong opinions) I felt the immediate desire to get an editable (2D) scene that I could rearrange. For example I often have a specific vantage point or composition in mind, which is fine to start from, but to tweak it and the elements, I'd like to edit it afterwards. To foray into 3D, I'd be wanting to rearrange the characters and direct them, as well as change the vantage point. Can it do that yet?

[javier123454321]

This is the conundrum of AI generated art. It will lower the barrier to entry for new artists to produce audiovisual content, but it will not lower the amount of effort required to make good art. If anything it will increase the effort, as it has to be excellent in order to get past the slop of base level drudge that is bound to fill up every single distribution channel.

[moralestapia]

Still three or four order of magnitudes cheaper and easier than to produce said video through traditional methods.

[nomel]

I think inpainting and "draw the label scene" type interfaces are the obvious future. Never thought I'd miss GauGAN [1].

https://www.youtube.com/watch?v=uNv7XBngmLY&t=25

[jstummbillig]

> A way to test this is to take a piece of footage or an image which is the ground truth, and test how much prompting and editing it takes to get the same or similar ground truth starting from scratch.

Sure, if you then do the same in reverse.

[mattigames]

Not too far in the future you will be able to drag and drop the position of the characters as well as the position of the camera, among other refiment tools.

[estebarb]

For those scenarios would be helpful a draft generation mode: 16 colors, 320x200...

[torginus]

Yeah, it almost feels like gambling - 'you're very close, just spend 20 more credits and you might get it right this time!'

[bilsbie]

Sounds like another way of saying a picture is worth a thousand words.