There are naysayers with every new technology but we always find ways to live with them and make them useful. I’m excited. Let’s do a survey of the magical land of AI and how it might apply to Hollywood.
I do believe that this technology, especially in combination with other strong tools such as Unreal Engine, will be important for everyone. It will be helpful for writers and it will radically reduce the cost of production for certain kinds of productions, especially FX-heavy shoots.
And bear in mind, these are the earliest days. What is interesting today, we should imagine will be much further along in six months. The studio and workflows of tomorrow definitely look different from the studio and workflows of last year.
STORYTELLING
Here is a completely AI-generated show notionally patterned after Seinfeld that you may have heard of that is continuously autogenerated and streamed to Twitch. It does not give real TV comedies a run for their money, but it is worth looking at if you bear in mind that technology usually gets better quickly.
Text AI apps can outline, can tell stories, can generate character descriptions. Right now, the weakness of ChaptGPT as a helper, at this point in time, is that it is like having a team of writers who are logical but they don’t really get comedy or your show. But they can come up with ideas that make basic sense. Sometimes if you’re writing a treatment or something, it’s easier to revise a not very important page (that appeared in ten seconds) than to start from zero.
Plotdot.ai - This is a specialized AI just for storytelling. This is getting more serious. It is at a very early stage but I think represents the next generation of where this will go. You sketch in a story, even very lightly, and it outputs a prose story of up to 100 pages in whatever genre and style you choose. If this could generate an outline and iterate based on detailed notes, it could actually be helpful. Obviously the good things about the story, the unique and original beats and things about the characters, will be the things you put in, but it’s not inconvenient to have the AI flesh out the treatment or raise issues about character descriptions. I actually suspect this will be most helpful for outlining and analyzing stories.
These will be more helpful as they begin to internalize some storytelling ideas and become more interactive and develop the ability to track themes and see 3 acts or eight sequences, etc. And I think it might be most useful when it gives feedback. A simple example would be if you kill the villain with a magic jade dagger in act three, it might remind you that you have to set up that dagger earlier in the story. I predict writers will widely adopt more robust future versions of these.
Character Design and Pre Viz
Of course, you can create characters and maintain them across sessions (though this should be more intuitive).
In Midjourney, I entered “zombies zombie heads by ffkz, in the style of tex avery, presentation of human form, quentin blake, alien worlds, gesture driven, stephen hillenburg, 1970-present --ar 16:9” and got this back.
Honestly? Not a bad start.
This can seem like a problem. Or it can seem like an opportunity.
The problem I sense creatively is that it can only harken back. It can’t come up with something groundbreakingly new. Can’t it? Maybe it can. Don’t count it out.
But this can’t be ignored because that command took me ten seconds.
At minimum, if you gave it a design and said “give me an expression sheet,” it could certainly do that. Or it will be able to.
If you fed it a scene and character designs, one imagines it will soon be able to spit out a storyboard that you could give notes on and iterate on. But one could imagine there might be a future app designed specifically for animation that is more original and more formidable.
Fine Art
I should not ignore the fact that there is a lot of very interesting work being done in straight up 1/1 editions of fine art, exemplified by Claire Silver, who has been working in AI for years and whose work you can see at LACMA. And to some extent this provides a counter-point to the issue of whether, in the right hands, and perhaps combined with other tools for post-processing, AI can only “harken back.”
“Parody” and Borrowing
AI is able to create very realistic, fake images of real people.
This extends to video and voice as well. Here’s an amazing AI generated video with the cast of the Matrix plus Wilford Brimley singing Ice Ice Baby.
Stallone as The Terminator:
Drake and The Weekend collaborating on a song they never collaborated on. At minimum, this creates the opportunity to cast 1984 Arnold Schwarzenegger in eight movies next year. Matt Stone and Trey Parker are working on that ability at their startup Deep VooDoo. Check out their Kendrick Lamar video —
Casting actors younger, casting Veronica Lake, it’s all an option now or it will be soon.
VFX and Production
Direct text to video AI is at a rudimentary stage. Here is an example.
The leaders now appear to be Runway ML and ModelScope but many are in the works.
People are producing digital comics and animation, where, in creating the animation, the AI is working extensively with other tools.
Digital comics example in detail:
A strong animation example from Corridor Digital:
Things are getting much, much easier on the VFX front. You can see some very interesting videos here and here. You can do things quickly on your laptop that used to require more sophisticated vfx teams. It seems reasonable to project that the budgets of complex vfx driven projects are going to come down considerably. AutoGPT can run locally and interact with other apps and APIs, which means it can perform tasks for you. One can imagine giving it a storyboard with instructions and having it manage a workflow across multiple apps to come up with a draft of a scene. With the advances in Unreal Engine and Unity, this will makes many things easier and cheaper.
Games are their own category that I am not going to delve into but I would note that ChatGPT can code and can already code up text adventure games. Creating game assets and game environments and situations with multitudinous realistic characters responding in real time seems quite realistic for stand alone games or games that supplement movies.
PreViz, Development and Movies That Don’t Exist Yet
Julie Design has created some interesting images and sequential art for movies that don’t exist (yet). Whether you are thinking through an idea or selling an idea or popularizing an idea, this seems useful. I actually love this process for thinking through and playing with ideas.
I’ve made some posters for ideas and projects, and I think it’s helpful in the development process. It can just help think something through. Is this what the project is?
Yes I would like to make a somewhat scary stop motion version of Animal Farm.
Cyberpunk heist.
I love the Douglas Sirk vibes for this period Hollywood soap.
Your control isn’t perfect though, at least at my level of experience. Sometimes you want hip, and you get more like “sultry perfume ad.”
Future Capabilities
Again, this is literally day one. You have to imagine all of this much, much better.
Legal Issues
Are there copyright issues if Large Language Models are reading existing works and then producing output based on their learnings?
Well. Not if they are merely “inspired” by the works and not outright copying. Which on the whole is indeed what is happening. But this will be an ongoing issue in visual arts, writing and music. Can Warner Bros. train a model on all of its animation produced to date? Do directors have a say in that? Does anyone else?
Separately, things created by computers don’t receive copyright in the United States. So … if an AI helps you create something, do you have the commercial rights? Can you claim copyright? You can if you modify it. But not if you don’t. For now.
On the plus side, it will likely create opportunities for actors and performers to license out there likeness and voice rights and I believe that CAA has already set up a group to manage this.
Related Developments
Crowdfunding, DAOs and Open Platforms
I wrote in December about How to Develop Film and TV Like It’s 2023, where I said that producers should test content on the web as part of development and try to attract fans and a community.
I would have to revise that now to include these AI tools. With AI, you will be able to more easily reify ideas and get them into the world in some form. Not in their final, finished form but in some form that gets a reaction and where you can get feedback. We have many open platforms such as YouTube, TikTok, wattpad, Webtoons, and others — even just Reddit or Twitter — where we can put new ideas in front of people and see if they get traction. We are no longer limited to just discussing ideas theoretically in conference rooms. We can create some manifestation of the idea and put it into the world, perhaps attracting a community of fans.
Added to this, crowdfunding platforms have grown and whether you use web3 or something like republic.co or startengine where you can sell shares in a media project, you should be able to raise money using the feedback and traction you have generated online. A show called The Chosen has raised over $25MM this way.
Finally, DAOs are online organizations that do things, like clubs. Why not a global culture maven DAO to help find and evaluate great ideas? Production companies are limited by the eyes and ears of the people who work there, usually 5 or 6 people in LA with their usually shared biases. But what if you had 3,000 people looking for ideas and helping evaluate ideas where you had a system for compensating those people in success? That could work. I think it should be called All the Fun Things.
It’s almost like you could manage an independent studio just with one person from your living room.
Maybe that’s asking too much. But I will not be surprised if the next major media thing comes from a smallish team, including a small, elite, group of artists, writers, vfx people and directors, where the development, testing and production processes are augmented and accelerated by AI, and where development is much more multi-media and public than it typically is today. That’s tomorrow’s Hollywood.
That’s not going to be the way for the next Birdman most likely, but the next Avengers, Star Wars, etc. I think can come from such a system. And maybe Birdman could be helped in some way, too.
This is either all exciting or all disconcerting or some mix of the two but I suspect you’ll find a way to make it your ally.
The best way to keep up is to follow some key people on Twitter. I would certainly include Robert Scoble, Dave Craige, Claire Silver and Bilawal Sidhu, but the easiest thing might be to follow this list Robert Scoble put together called AI/ML.
RP
“This is either all exciting or all disconcerting or some mix of the two but I suspect you’ll find a way to make it your ally”
For those on the disconcerting side of the ledger. How about another piece about all the thing that could go wrong and the havoc that could be unleashed... hey might be a film there somewhere?