AI in Hollywood Today : “Not a single frame untouched”

AI in Hollywood

There is a growing resistance to AI in Hollywood.. and AI-generated content flooding the internets. The general concept of this genAI content flood has been dubbed “Slop”… images, text, and videos that are laser-targeted to the algorithm, that appear at first glance to be some of the [coolest architecture / most beautiful faces / most awesome nature destinations]… but which in the end, are total AI-synthesized pixels that are the net equivalent of empty calories at best, and sickly sweet sugar candy that make your teeth fall out, at worst.

One idea, which we are all in fact witnessing first hand in both our search results and social media feeds, is a rapidly growing flood of AI content which “clogs up” and pushes out the human-created “natural” media. Two challenges are that a) the content actually looks better at a glance, and b) it is better suited to the algorithm, down to the individual pixel and caption.

AI in Hollywood - AI Slop clogs our search results - Baby Peacock

 

.

A third almost spookier component is that the AI has learned (expertly) to make photos and videos that look amateurish… shaky cameras, poor lighting, camera-too-close, noisy backgrounds… so for those of us who have ascended to the next level and have tuned our visual filters to sort out the human-gen from the Slop… well, that too has been compromised.

In response to all this, various formal and informal groups have sprung up, who are championing the idea of “100% organic human-created content, utterly free of AI meddling.” There is a critical flaw to this thinking tho, and that is this: There is barely a single production tool today (think PhotoShop, Premiere, Davinci, Avid, Final Cut Pro, Ableton, Logic Pro, Pro Tools, etc) that does not have AI already deeply integrated into its codebase, toolset, and workflow.

In other words, while we may be sensitive, even allergic to 100% AI genn’ed content, (and soon, I promise, you will not be able to tell the difference!)… for the past 5 years, there is barely a single frame of any movie or TV show that we’ve seen, that has been left untouched by AI processing algorithms. Most of these are quite subtle, and might not riase hackles. They are largely tasks that would have been done by human technicians, at 100x the cost and 1,000x the time. For instance:

  • wire removal
  • color grading / correction
  • upscaling
  • rotoscoping

But then we get to uses that were financially unfeasible or technically impossible prior to AI tools, such as:

  • vubbing (voice dubbing)
    • example 1 : change the dialog (script edits post-shooting)
    • example 2 : auto-translate footage into any/all languages, with accurate re-rendering of lip movements and facial expressions.
  • de-aging, re-aging
    • this has been done in the past, by human VFX artists, most memorably with Benjamin Button… but we are about to witness next level this christmas with Tom Hanks and Robin Wright in Here (Nov 2024)

  • digital doubles
    • this is one of the Holy Grails of VFX, which is finally being realised with the deployment of AI tools.

 

[DRAFT 0.3]

In the Age of AI SmartPhone Cameras, what is Real?

PHOTOGRAPHIC EVIDENCE / THE END OF TRUTH /
HUMANITY ENTERS THE AGE OF LIVING MYTHOLOGY

Meta’s new “Movie Gen” AI system (and competitors) can deepfake video from a single photo

A future where realistic video fakes are easy and plentiful seems inevitable.

WHAT THIS MEANS:

(a conversation with Abram Hodgens)

GR (quoting and reacting to article): “It’s bringing us closer to a cultural singularity where truth and fiction in media are indistinguishable without deeper context due to fluid and eventually real-time AI media synthesis”

I would use the word “apocalypse” or “crisis” rather than singularity, but thats just semantics. We had ~50 years of confidence that “photographic evidence” (later, video footage) was the bedrock of historical truth. Now all of that is completely up for grabs. Its all muddy. All fantasies, all alternate historical timelines, and all conspiracy theories, will have synthesized video evidence that will be totally indistinguishable from “real footage”.

This has already happened with smartphone cameras. Snap (filters) started it, but the latest versions of both Apple iOS and Android don’t leave a single pixel of your photos untouched by AI. background / object removal, expression modification (make johnny smile), etc. its all fluid. the “ground truth” is MIA.

Apple stated :

we are not trying to be true to the pixels
coming from the image sensor.
we are trying to remain true to
the emotional memory of the user
the internal story.”

— WUT???

AH: “can anyone here see any benefits of this technology?”

GR: I can see one primary benefit. The total democratization of film-making. There are many humans who have a “movie in their head” that they want to make. Now still, >97% of them are never going to make their movies. But there is a segment, who could never afford the production, or never get their script bought, who will make novel, interesting, moving AI films. I am looking forward to viewing the best of them.

AH: i’m not so convinced that “poor ol’ farmer Joe in Iowa and farmer Nguyen in Danang” being able to tell stories with video is adding that  much value, especially compared to the potential downsides… one could argue that ugly girls using snapchat filters to become pretty is of value too, but the downsides are far greater than any benefit.

GR: Agree 100 that downsides outweigh upsides. also keep coming back to android jones > Altman: “so, you are building this super-intelligent AGI that supposedly will cure cancer and solve climate change, but first… lets basically put every commercial artist on the planet out of work… HOW is that part of the plan? WHAT is the value to humanity?”

 

SotA Video Gen models (2024-10):

.

Where it’s Heading: Deep Deep Deepfakes

.

Additional (future) expansions:

  1. the dynamically targeted and auto-genn’ed personalized video ad on insta
  2. the relativity of truth and the re-birth of mythology
  3. reference Yves Bergquist (USC ETC) from theWrap theGRILL conference
    (“AI in Hollywood” Executive Summit at DGA Theatre, Oct 8, 2024)
  4. If Digital Doubles can do the work (unsupervised), is there actually an Identity (person) behind the Celebrity Image, or is it simply an empty shell, a cartoon, a meme and a trope for society at large?
  5. What happens when (see Japan) the first fully synthetic superstars take the stage, and amass huge fan bases? And enter the influencer leaderboards on YouTube, Insta, TikTok? The corporations and artists will “own” and profit from these “digital slaves.”

.


prompt: “a glowing blue pixelated holograph plane hovering in front of a user holding a smartphone, taking a picture of a natural landscape on a soundstage.”

engine: MidJourney v6.1 Alpha