OpenAI has developed a new text-to-video model called "Sora". Soon anyone will be able to script, direct, and create their own films, their own cinematographic works.

The chair video feels like an SCP. I love that they included some failed generations for some nightmare fuel.

This is such a great example of the impressiveness and flaws of this tech.

Look at the weird non corporal chair everyone, might forget the people interacting with it are ai generated too
Imagine if the AI can add sound to the video, it would be fucking nuts.
- It wouldn't be too hard to train. There are enough audio models and computer vision models that could be trained in parallel on video clips that have recorded sound to train what sound profiles are associated with what events in the frame.
  
  The real fun one would be to figure out how to train an AI to understand sounds originating from out of frame.
- It won't be long before we have something like Oobabooga or Stable Diffusion but for artificial video with matching audio. I'm so sorry for historians in the future trying to determine whether this video of Joe Biden doing tap dances on an F-16 while throwing a hadouken recovered from a trashed hard drive is authentic or not.

There's more to making movies than generating moving images.

Of course, but remember where the tech was a year ago? Where would it be in a year from now?
- Off the chart porn with tay tay on a plane I believe is the correct answer to this.
Yeah, they'd have to generate text, music and speech as well -- who ever heard of something like that?

Anyone = anyone with our premium subscription, subject to regional availability.

Porn, fake political videos, and avoid paying for stock clips.

These are the main use-cases.

Also, killing time while waiting for the edibles to wear off.

The film industry could use some of the tools, like assisting in special effects with reduced costs. As long as it doesn't claw on the industry workers with an abusive contract, that is...

As someone who has studied cinematography, I'm not worried about tools like this. Visual storytelling requires intention that AI can't provide. This will probably be fine for improv comedy with simple reaction shots, but those tend to lack photographic intent anyway.

Films like Children of Men and Drive, that tell their stories through visuals more than exposition, will still require a trained eye to craft for the foreseeable future.

I mostly agree with you, except:

Visual storytelling requires intention that AI can't provide.

It should be "that AI can't provide yet". The intent can be provided via elaborate prompts. It is just that the output from current generative AI isn't up to that level yet.

Give it 5 years, and AI might be able to do what you mentioned.
- I don't think an AI will ever achieve new things like the dolly zoom or bullet time. AI can replicate these things once they already exist, but humans are brilliantly absurd and we make strange new art all the time.
A false sense of security is worse than no security at all.

When the AI workforce breaks through the arts and film industry, what makes you think you're the one holding the keys and calling the shots on AI? Simply because you studied cinematography?

The point is the scale of the impact will likely be substantial and many potentially will be displaced. I would argue you should be studying this shift rigorously instead of being dismissive to give yourself the advantage.
- Idk, they can't stop hallucinating extra fingers yet. I've been running stable diffusion and llama locally for a while. I'm not worried about cinematography for the near future. And you're being a fear mongering dick. Bloooocked!

Maybe now we can get a good ending to Game of Thrones...

No, soon all AI will be paywalled and feature locked. Right now we are just getting a free taste so we get hooked while they refine it. There is no way capitalism will let us have so much power as individuals.

Hopefully FOSS can catch up or become even better than what they have.

So if you want a movie where every solid object in every scene is morphing, dripping, and twitching - you're in luck, because ChatGPT can produce them in 10 second bouts.

Want something with characters interacting? Moving? Or with their lips moving when they talk? GO SOMEWHERE ELSE BUDDY - because these models can't do that.

Want to narrate a series of morphing AI images and pretend it's a "movie" - great, COME ON DOWN!!!

Hey, the first airplane sucked. No pressured cabin or snacks with the Wright brothers was there? We’ll just have to wait and see how it improves with time
- Honestly, I saw some of the videos of what SORA can do, and deleted my old comment - as it was based on what the previous video models were producing.
  
  It still creates stuff that isn't realistic (eg. Getting proportions wrong, or erasing objects when they're obscured), but has certainly come a long way.

THIS IS FUCKING MADNESS

No way this backfires.

Pr0n, once again

It was only a matter of time and I’m excited for it

The guy that stole the last dlc spot in Smash Bros?