Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Meta Movie Gen

ai.meta.com

Post reply on HN
Print view View original

381–390 of 1001 posts

Re: Meta Movie Gen

#381

Post by halyax7 »

adventured wrote: post #185
dpcan wrote: post #111

You know it’s going to happen: “I want a funny road trip movie staring Jim Carey and Chris Farley, based in Europe, in the fall, where they have to rescue their mom played by Lucille ball from making the mistake of marrying a character played by an older Steve Martin.” 10 minutes later your movie is generated. If you like it, you save it, share it, etc. You have a queue of movies shared by your friends that they like…

It'll require tens to hundreds of hours to script the flow of the AI content, to edit, make adjustments, clean it up, make the scenes link together smoothly, fix small glitches. Even with far more advanced AI, it won't come together like a movie people would enjoy watching, without vast human labor involved. Sub one percent of people are going to be willing to put in the hours to do it. The bulk of the spammed create…

Much of those editing steps could be streamlined and/or straight up automated so that estimate will come way down over time

Re: Meta Movie Gen

#382

Post by causal »

diggan wrote: post #360
causal wrote: post #162

Earlier quoted context omitted.

They have released a ton of open source? Llama 3 includes open training code, datasets, and models. Not to mention open-sourcing the foundation of most AI research today, pytorch.

Llama 3 is licensed under "Llama 3 Community License Agreement" which includes restrictions on usage, clearly not "Open Source" as we traditionally know it. Just because pytorch is Open Source doesn't mean everything Meta AI releases is Open Source, not sure how that would make sense. Datasets for Llama 3 is "A new mix of publicly available online data.", not exactly open or even very descriptive. That could be anyth…

Sure, the Llama 3 Community License agreement isn't one of the standard open licenses and sucks that you can't use it for free if you're an entity the size of Google.

Here is the Llama source code, you can start training more epochs with it today if you like: https://github.com/meta-llama/llama3/blob/main/llama/model.p...

It's rumored Llama 3 used FineWeb, but you're right that they at least haven't been transparent about that: https://huggingface.co/datasets/HuggingFaceFW/fineweb

For models I prefer the term "open weight", but to assert they haven't open sourced models at all is plainly incorrect.

Re: Meta Movie Gen

#383

Post by sangnoir »

mywacaday wrote: post #54
cs702 wrote: post #35

Impressive. It's only going to get better, faster, cheaper, easier.[a] Sooner than anyone could have expected, we'll be able to ask the machines: "Turn this book into a two-hour movie with the likeness of [your favorite actor/actress] in the lead role." Sooner than anyone could have expected, we'll be able to have immersive VR experiences that are crafted to each person. Sooner than anyone could have expected, we won…

Are we only a few years away from one person/small group made movies where the dialog, acting, location and special effects can be tweaked endlessly for a relatively low cost. If I was a studio exec I'd be worried.

> If I was a studio exec I'd be worried

Counterpoint: home "studio" recording has been feasible for decades, but music execs are not ruffled. Sure, you get a Billie Eilish debut album once a generation, but the other 99.99% of charting music is from the old guard. The media/entertainment machine is so much bigger than just creating raw material.

Re: Meta Movie Gen

#384

Post by tyjen »

While we're still a fair distance away from creating polished products capable of replacing Hollywood gatekeeping; the bursting of the creative dam is on the horizon and it's exciting! I'm looking forward to when you can write a script and effectively make your own series or movie. Tweaking it as you go to fit your vision without the exhausting a large amount of resources, capital, and human networking to produce similar products pre-AI.

Re: Meta Movie Gen

#385

Post by blueblisters »

jsheard wrote: post #81
dageshi wrote: post #33

Earlier quoted context omitted.

Yeah but... it's good enough? There were movies with horrible VFX that still sold perfectly well at the time.

An important contrast is that early VFX offered strong control with weak fidelity, and these prompt-based AI systems offer high fidelity with weak control. Intent matters if you want to make something more than a tech demo or throwaway B-roll and you can't communicate much intent in a 30 word prompt, assuming the model even follows the prompt accurately.

Yeah controlnet-style conditioning doesn't solve for consistent assets, or lighting, framing etc. Maybe its early but it seems hard to get around traditional 3D assets + rendering, at least for serious use-cases.

These models do seem like they could be great photorealism/stylization shaders. And they are also pretty good at stuff like realistic explosions, fluid renders etc. That stuff is really hard with CG.

Re: Meta Movie Gen

#386

Post by bee_rider »

tsimionescu wrote: post #374
intended wrote: post #332

Man, everyone is happy with these advancements, and they are impressive. I’m here looking at users and wondering - the content pipelines are broader, but the exit points of attention and human brains are constant. How the heck are you supposed to know if your content is valid? During a recent apple event, someone on YT had an AI generated video of Tim Cook announcing a crypto collaboration; it had a 100k users before…

We are at a crossroads of technology, where we're still used to the idea that audio and video are decent proof that something happened, in a way in which we don't generally trust written descriptions of an event. Generative AI will be a significant problem for a while, but this assumption that audio/video is inherently trustable will relatively soon (in the grand scheme of things) go away, and we'll return to the his…

That’s an interesting thought.

I think I would not like to live in a world in which democracy isn’t the predominant form of government. The ability of the typical person to understand and form their own opinions about the world is quite important to democracy, and journalism does help with that. But I guess the modern version of image and video heavy journalism wasn’t the only thing we had the whole time; even as recent as the 90’s (I’m pretty sure; I was just a kid), newspapers were a major source. And somehow America was invented before photojournalism, but of course that form of democracy would be hard for us to recognize nowadays…

It is only when we got these portable video screens that stuff like YouTube and TikTok became really important news sources (for better or worse; worse I would say). And anyway, people already manage to take misleading or out of context videos, so it isn’t like the situation is very good.

Maybe AI video will be a blessing in disguise. At some point we’ll have to give up on believe something just because we saw it. I guess we’ll have to rely on people attesting to information, that sort of thing. With modern cryptography I guess we could do that fairly well.

Edit: Another way of looking at it: basically no modern journalist or politician has a reputation better than an inanimate object, a photos or video. That’s a really bizarre situation! We’re used to consulting people on hard decisions, right? Not figuring out everything by direct observation.

Re: Meta Movie Gen

#387

Post by dguest »

tsimionescu wrote: post #374
intended wrote: post #332

Man, everyone is happy with these advancements, and they are impressive. I’m here looking at users and wondering - the content pipelines are broader, but the exit points of attention and human brains are constant. How the heck are you supposed to know if your content is valid? During a recent apple event, someone on YT had an AI generated video of Tim Cook announcing a crypto collaboration; it had a 100k users before…

We are at a crossroads of technology, where we're still used to the idea that audio and video are decent proof that something happened, in a way in which we don't generally trust written descriptions of an event. Generative AI will be a significant problem for a while, but this assumption that audio/video is inherently trustable will relatively soon (in the grand scheme of things) go away, and we'll return to the his…

It's not all the way back as long as solid encryption exists: Tim Cook could digitally sign his announcements, and assuming we can establish his signature (we had signatures and stamps 200 years ago) video proof still works.

So we're not going all the way back, but the era of believing strangers because they have photographic or video proof is drawing to a close.

Re: Meta Movie Gen

#388

Post by sangnoir »

indigoabstract wrote: post #137
cs702 wrote: post #35

Impressive. It's only going to get better, faster, cheaper, easier.[a] Sooner than anyone could have expected, we'll be able to ask the machines: "Turn this book into a two-hour movie with the likeness of [your favorite actor/actress] in the lead role." Sooner than anyone could have expected, we'll be able to have immersive VR experiences that are crafted to each person. Sooner than anyone could have expected, we won…

It looks impressive , yet I'm not feeling very impressed. If only I could get as high as you do from watching those demo reels. Out of curiosity, what is it that people do with these things? Do they put them on TikTok?

> It looks impressive, yet I'm not feeling very impressed.

The demos was made by nerds (said with love) with a limited time window. Wait until the creatives get a hold of the tool.

Re: Meta Movie Gen

#389

Post by grumbel »

FileSorter wrote: post #151
turblety wrote: post #127

Why do these video generation ones never become usable to the public. Is it just they had to create millions of videos and cherry pick only a handful of decent generations? Or is it just so expensive there's no business model for it? My mind instantly assumes it a money thing and they're just wanting to charge millions for it, therefore out of reach for the general public. But then with Meta's whole stance on open ai…

There are usable ones runwayml.com pika.art hailuoai.com

klingai.com

lumalabs.ai

Post reply on HN