TurboDiffusion: 100–200× Acceleration for Video Diffusion Models (opens in new tab)

I think video-based world models like Genie 2 will happen and that they'll be shrunken down for consumer hardware (the only place they're practical).

They'll have player input controls, obviously, but they'll also be fed ControlNets for things like level layout, enemy placement, and game loop events. This will make them highly controllable and persistent.

When that happens, and when it gets good, it'll take over as the dominant type of game "engine".

A very, very different mechanism that "just" displays the scene as the author explictly and manually drew it, and yet has to pull an ungodly amount of hacks to make that viable and fast enough, resulting in a far from realistic rendition...

This on the other hand happily pretends to match any kind of realism requested like a skilled painter would, with the tradeoff mainly being control and artistic errors.

Bob Ross did it, too.

> If Sora/Kling/whatever ran cool locally 24/7 at 60FPS, would anyone ever build a UI?

I like my buttons to stay where I left them.

Please no, please no

That will be Windows 12 and perhaps 2 generations in of iOS :)

Nevertheless it does seem that generating will fairly soon become fast enough to extend a video clip in realtime. Autoregressive by the second. Integrated with a multi modal input model you would be very close to an AI avatar that would be extremely compelling.

I'm creating a college / corporate seminar level class with a 3D animated host and instructor. Each lesson begins with a brief animated intro, which the student then reads their lesson afterwards, and then there is a chatbot that understands that lesson that engages with the student afterwards. The course will have about an hour of 1-2 minute videos in the end, and because they are an animated "professor" it is then possible to create other ethic versions, and other language versions easier than otherwise. And for the curious, that hour of final in-use video will be sourced out of somewhere around 8 hours of final video; these video AIs are fine and dandy for short content, but when working on longer form media the consistency issues grow to Godzilla sized and really become the most consistent issue: trying to keep the likenesses of the characters, and the environment from drifting over time.

Also interested, since that is my same impression.

I tend to think that it leads to censorship, and then censorship at a broader level in the name of protecting our kids. See with social networks where you now have to give your ID card to protect kids.

The best way in that case is education of the kids / people and automatically flag potentially harmful / disgusting content and let the owner of the device set-up the level of filtering he wants.

Like with LLMs they should be somewhat neutral in default mode but they should never refuse a request if user asks.

Otherwise the line between technology provider and content moderator is too blurry, and tomorrow SV people are going to abuse of that power (or be coerced by money or politics).

At a person / parent level, time limits (like you can do with web filtering device for TikTok), content policy would solve and taking time to spend with the kids as much as possible and to talk to them so they don’t become dumber and dumber due to short videos.

But totally opposed that it should be done on public policy level: “now you have right to watch pornography but only after you give ID to prove you are adult” (this is already the case in France for example)

It can quickly become: “now to watch / generate controversial content, you have to ID”

You just fell outrage bait so I doubt you will be able to identify AI.