◡◶▿ SOFT12 | Future legends
🗿 Pata-mythological tropes & cyberpunk filmmaking gigs. Plus: rare screening of Mr Cole's final movie; 10,000 imaginary diarists; end of time. | Imaginary Software of the Filmmaking Future Week 12
Hello. It’s the final week of our spring term module: Imaginary Software of the Filmmaking Future. By the time you have finished this email, you will be ready for the future. Which is perfect, because that’s when the future begins.
This post also marks one year of uninterrupted weekly UPV dispatches. We have learned so much about filmmaking. And yet, we still know so little! Possibly less than before. But at least we’re happy. Right?
Missed a week? Joined late? Don’t worry about reading these lessons out of order. Each functions independently. They are sent in a sensible sequence but hardly reliant on it.
Quick recap! Last week, we covered how:
Directors don’t know much about acting, and actors know too much, and this seems to balance out quite well.
However, the potential to replace human actors with digital sprites threatens this balance.
Crooked talent agents, jaded ex-Soviet bloc gym trainers, and out-of-work actors will be queueing up to get their hands on the joypads of tomorrow’s synthetic cinema stars.
Technicians and the actors whose former faces they control could develop seething rivalries - and this could breathe new life into the tired film industry gossip columns.
I also mentioned how, with more time1, we would expand the lesson out of ‘AI actors’ and into ‘AI characters.’ I will instead signpost you to
’s post on the practicalities of basic, large-scale character fabrication. Or character ‘seeding,’ since each set of character ingredients is just the start: the personality of each of Gioia’s 10,000 characters2 develops ‘under the hood’ through lightly shepherded machine entropy.Gioia’s work also hints at the practical steps necessary to realise Volodymyr Nanneman’s device for generating 65,536 unique character types. And Francis Dove’s Undepth In Real People And Those Who Believe They’re Real. Would’ve been great to cover Nanneman’s work more closely in our present module. Read these↑ primers instead! Read Mr Gioia’s thing!
Shall we do this week’s lesson? Okay. It’s a nice long read to finish up the term. But remember, like every week, each micro-essay below works as a standalone piece. So you may dip in and out as your study schedule allows.
In today’s lesson, let’s look at how things might shape up in an AI filmmaking landscape. The future shape of:
🧖 The human, who may be replaced by charismatic goop.
💪 Job(s), which may become rare, alienating, and perilously cool.
🎞️ Movie grammar, which will eat and regurgitate itself.
🧠 And the human mind - ‘that old thing.’
We’ve talked several times about how the human mind is the greatest software of all. What do you think now we’re at the end of the program? Is the human mind still the best? Maybe computer software is better. 🤷♂️.
Scroll on for all the learning. And for details of a rare screening of my “career-ending” movie (among undoubtedly better movies) in Italy next week. And some bursaries for young people.
End of the human form
You can hear me deliver this lesson by scrolling up to the header and clicking Listen and/or the play ▸ button.
The ability to use AI to create convincing synthetic photorealism need not mean the end just of actors, but rather more broadly of the human form as fetish object altogether.
Character designers can go beyond The Lion King, or even - even! - Avatar to create new beings based on animal morphology or, more likely, otherwise.
What do you imagine when considering this potential? Some kind of wisecracking goop? Imagine how much better the character forms imagined by the computer might be!
Early experiments with AI-generated video have revealed that, in fact, goopy, mutant character morphology is an advantage. An advantage, because video generation is a mutant process. It starts with a prompt or key frame which evolves in unpredictable and counterintuitive ways, creating blobs and pulsing tics.
Yes, blob-like characters and corrupt worlds look best. A thriving, secret film industry that’s been bubbling away beneath the surface all this time, waiting to be released. No more humans. Let the digitised collective unconscious find its own weird form.
3D printing future films
There might be a computer that can do everything except for making moving images. A computer with no photographic or animation facility.
Instead, it would cook up everything for you: plot, people, lighting, sound, schematics, props, special effects. And output it to the physical world as holograms, laser beams, 3D printed animatronics, or any combination of all of those and more. It would be up to the filmmaker to photograph (film or video) the movie as it unfolds before their eyes.
Of course, the filmmaker would have little freedom to get arty with angles or close-ups. Take the wrong angle, and you would expose the edges of the movie or film the rear workings of its sets and characters; go too close up on a synthetic prop or face intended for a wide shot, and you’ll reveal unpalatable levels of surface grain or manufacturing flaws.
But this computer would save a lot of thought. A lot of work. And enable the filmmaker to spend a lot more time dreaming up photographic ideas with her team of cinematographers.
Video training
Early AI video generation tools were trained on pre-existing moving images. Fed thousands of hours of footage to learn how movies should move, how the pictures should look. These tools continue to gobble up emerging video materials, which are themselves increasingly AI-generated.
To prevent the phenomena and visual artefacts of ’inbreeding’ due to a lack of ‘real’ (photographic) contemporary training footage, a new, ongoing task needs to be fulfilled. That task is to create new stock footage with which to train future filmmaking machines.
A job making IRL movies for AI training will likely be thrilling and banal. Thankless - a shadow task - and handsomely paid.
The filmmaker of training materials will work for an agency and spend her days fulfilling the shots they have arranged. Her footage will be valued for its unremarkable quality: generic, comprehensive, the showroom dummy films of live-action shooting. Her agents and clients will grow tired of acknowledging that such work requires great skill. She won’t mind.
She might mind that they fail to recognise that despite their desk-bound production work, she has to do most of the fixing, arranging, and re-arranging of shoots on the road.
After a shoot, she will catalogue and submit her work via satellite, choosing to ignore the politics of corporate espionage and the threat of clip-jacking. There are dozens or even hundreds with her job worldwide, filing training footage for rival AI firms. She reads the industry journal on paper to keep her sense of depth and texture fresh, or perhaps to evade eye trackers.
Her training footage needs to be fresh and up-to-date because the nature of image quality, content (character fashion, gestures, architecture, vehicles, meteorology, wildlife), and ways of looking change all the time. Her client’s software wants to be ahead of the game.
For a while, AI-generated footage will likely follow the conventions of the camera arts. The limits of photography when bound by physical apparatus. That’s because it has been trained on photographic materials. And because audiences seek familiar forms.
But AI video needn’t appear to be created with a camera. The influence of the camera may eventually fade. And the filmmaker whose job it is to create raw training materials for the AI may benefit from developing uncameralike camera techniques. Cigarette hanging from lip as body dangles from man-sized drone. She is the last living filmmaker and the first to see the world through tomorrow’s eyes.
Pata-mythological tropes
In 1967, Italo Calvino wrote about the possibility of a literature-generating machine. He began by mentioning the interchangeable nature of the blocky, archetypal elements of primitive folk tales.
“The storyteller explored the possibilities implied in his own language by combining and changing the permutations of the figures and actions, and of the objects on which these actions could be brought to bear,” Calvino wrote of the earliest tribal tales.
“What emerged were stories, straightforward constructions that always contained correspondences or contraries - sky and earth, fire and water, animals that fly and those that dig burrows; and each figure had its array of attributes and a repertory of its own.”
Calvino imagined a machine that might reassemble and juxtapose these units into mechanical stories. Decades on, the machines are certainly giving it a go. And they are applying this logic to moviemaking, too. However, any movie-generating machine must go beyond what can be expressed in words or the actions taken by more or less archetypal figures.
The filmmaker’s figures are complex, multi-dimensional configurations of the superstar-actor-character. And these odd figures and their actions (which include more or less scripted gestures, expressions, and continuity errors) are not the only interchangeable units. The sense(s) of a movie are built also from innumerable variable phenomena, and from the qualities of the moving image and sound and audiovision. Such as:
Moving visual detail (props, meteorology, changes in colour and light,
and the interplay between and beyond things like these),
camera movement (or more properly, the mutation of the image volume),
certain cuts and transitions,
sequencing and rhythms,
location,
location bleed3.
These are among the building blocks of a movie. The legends that populate it and give it sense, meaning, and feeling (and where necessary, narrative) as much as the figures and their actions.
In the photographic cinema, filmmakers reimagine and recreate these blocks from one movie to the next. A dusty shaft of light here, a scream there. A quality of surface grain or the hurried framing of a scene shot just for exposition. Just as storytellers thread the hunter or the rabbit through the folklore corpus.
But for an AI-generated movie, the filmmaker draws from an atomised library of movies. The original blocks are disintegrated and their particles tagged and filed according to the logic of the machine and its janitors.
And so, the new legends of cinema will be assembled from recurrent atoms - bits of girls and guns, cigarette smoke, certain hues or vocal tones - across the emerging canon of software movies.
They may resemble the ‘legends’ of cinema as we know it. Resemble common visual and audio and audiovisual ideas that have made it into permutations (hit movies) across a broad (multi-textural) but shallow (short) history of cinema.
But here, these tropes are reconstituted according to machine logic. They will differ from the originals in repetitive ways. And be joined (and perhaps overpowered) by:
Weirdly reconstituted legends that hadn’t been recognisable as legends during cinema’s pre-AI history, but which now become obvious.
Dislodged from the 20th-century subconscious by blunt 21st-century machinery.
New legends conjured by the machine’s peculiar preferences for certain permutations of movie-atoms.
We can expect to see this pata-mythological effect - the trans-dimensional regurgitation of cinematic tropes on a material level - refract through the future canon.
‘Clues’ and cliches re-encrypt themselves in bizarre new forms of film grammar or audio-image in subsequent movies. Particularly as future databases are replenished not with live-action footage but with the offal of the growing AI archive. The new bastard legends of a mutant lineage of neo-primitives.
The uncanny diner
Filmmaking software reshapes your relationship with the possible. With cause and effect. The world looks different when emerging from a session on the computer. Your universe and its logic shrink during use and take a while to pop back into shape after logging off.
It is something like the uncanny valley, to emerge from a conversation with an AI to have a conversation with a human. But it is not the artificial space of the AI conversation that feels off. It is your real life. Logging off from an extended software session is a step from the uncanny valley into the uncanny diner.
Creativity is like this more broadly. Quit making films, and your inner pitch generator will shrivel. The bit of your mind that deals with your next pursuit will gain influence the more you use it. Take up tennis and map your world in courts. Become a dog trainer and experience each non-work encounter from the perspective of a daft puppy.
The mind adjusts to the creative pressures you apply. How can you refresh? Reboot? Yawn or swallow until your ears adjust to the alien pressure? Do you even want your mind to pop back to how it was at ground level?
As we reshape our intellects and instincts, we must be wary not to let the ones we value atrophy. That’s what people say. Funny how the people all say that, these days. Funny how that’s what everyone’s mind comes up with.
Please share your thoughts, queries, and exercises from this week’s lesson in the comments.
Just when I thought I got out
A film I wrote and directed will play in Palermo, Italy, at Keja Ho Kramer’s Cadavre Exquis event on the 20th-21st April, 2024.
We shot UNIVERSAL EAR: The Curse of the Phantom Tympanum in the autumn of 2017 in Bourges, central France. And here is how we described it:
Time-travelling record producer Harley Byrne crash-lands in a virtual reality heritage theme park in 22nd-century France. Corrupt holograms, cyborg saints, and sentient statues haunt an absurdist Super-8 universe, digitally re-colourized for your pleasure!
The film played at IndieLisboa and Kinoskop, two wonderful festivals, among a handful of others, before being gobbled up by lockdowns and my growing antipathy towards making festival submissions.
Cadavre Exquis is programmed through a chain of recommendations - each filmmaker nominating the next. But around 20% of the films on show this year were made by someone with a variation on the name ‘Peter,’ suggesting that the system has gone wrong somewhere.
Also on this year’s program, you’ll find the work of my colleagues Aleksandra Niemczyk, Petr Makaj, and Peter Treherne, at least one of whom is my wife and several of which are called Peter.
Go along if you can!
I know some of you students are younger. And some are older! “Eternal.” Here’s something for the younger ones, or the children of the eternals:
There are two full bursaries available “for 15-17 year old global majority (non-white) females who would otherwise encounter a financial barrier to access and who are passionate about film & television” to join the Young Film Academy’s Filmmaking & Screen Acting Summer Camp this year. Global applications are possible for this UK camp.
Dame Judi Dench says, “Have a lovely time!”
Sounds like a life-changer! Don’t do it! Do it! Don’t do it!
It’s been a long semester, dear students. Can you believe we started in January? But unlike academia, filmmaking doesn’t tolerate long gaps. Unless you’re Terrence Malick or Victor Erice. Let’s meet here for a brief chat next Monday. Summer term begins in two week’s time.
Class dismissed!
~Graeme Cole.
(Principal)
📹 Unfound Peoples Videotechnic | Cloud-based filmmaking thought. ☁️
🐦 Twitter | ⏰ TikTok | 📸 Instagram | 😐 Facebook | 🎞️ Letterboxd | 🌐 Website
(or, say, a book deal)
Gioia’s characters were hired to create 10,000 unique personal diary entries. Each of which would make a fine cold open to an episode of Quantum Leap or perhaps better Person of Interest.
Location bleed: the influence of the ‘real daily life’ of a shoot location on the material). How do the expressions of passersby who didn’t even sign a release form contribute to the artificial filmmaker’s creative subconscious? or the sound of an unscripted dog barking, unscripted, in the woods in this movie and that?