Building a Tower of Babble

Remember when they tried to build a tower to reach the heavens and then they couldn't understand each other anymore?

Perhaps you saw this article?

Variety: Lionsgate Boss says AI can be used to adjust a movie's tone

‘Do It Anime, Make It PG-13. Three Hours Later, I’ll Have the Movie’

Hell yes we can do that, but that’s only the beginning.

I’ve been experimenting with the amazing Wan model and all the incredible tools that the community of open source models has been building around it for the past month or so. It’s hard for a traditionally-minded filmmaker to get one’s head around what it makes possible. Note that none of this is usable in films and television yet, the pixel-level image quality is still not there but you could get away with it for a dream sequence, or in a flurry of shots in a action sequence and you could definitely use it for social media and those ‘verticals’ I keep hearing about (and —embarrassingly— occasionally catch myself watching).

Here’s some simple things we can already do:

Take any character and animate them with full photorealism, have them say anything, do anything.

Make a character younger, older, change their performance, change them into someone else.

Change palettes, styles, the music in the background

Generate seamless transitions between shots, between takes.

Change the setting, change the lighting, the wardrobe, the language

You can see why we post people love and are enthralled by these possibilities, but of course they come at a cost…

Famously, when shooting ‘The Godfather’ Gordon Willis deliberately underexposed the negative, in order to ensure the studio wouldn’t be able to brighten up his now legendary top-lit chiaroscuro.

The film industry is rife with stories of studio interference, auteurs, sales and marketing people each with their own creative agendas attempting to seize control with hundreds of millions seemingly at stake.

Right now, controllable generative video works on a shot-by-shot basis and it is slow (and thereby expensive). We’re basically working in 80-120 frame chunks. It’s painstakingly slow to generate these shots (is it even filmmaking anymore?), however models that generate entire movies at once are probably just a year or two away at the pace at which we’re going.

Create a cool story and accumulate some core data (maybe some key performances, the main story points, some thematic background), then start versioning. There’s the festival version, the US-TV version, the theatrical version, the red band version, the director’s cut, the collector’s cut, the Senegalese version, the many varied game versions, the simulation, the one with you in it, the version that is set in the Star Trek universe…. and on and on and on. Infinite versioning crossover potential. So much babble.

So post will swallow up production and eventually the studio (or app) will swallow up post, as we all venture into the cloud.

I am already mourning the end of shared experiences, if all content is custom versioned then will the power of stories to connect us be negated?

Is this our Tower of Babble?