Show HN: Netflix for AI-Generated Videos
lucidbox.net
lucidbox.net
I'm imagining some kind of TikTok like algorithmic magic combined with this AI generated stuff is going to be a huge thing in the near future.
We removed the wrong additive...
Young kids already have access to the internet at a young age and that's screwing them up royally anyway
Plenty of stuff is already restricted as "adults only", it wouldn't be unprecedented to also restrict this. I can imagine various arguments for it - for example, "since it's generated by AI nobody can be held accountable if the content is inappropriate for children and there is no way to ensure the AI adheres to content restrictions, so it can't be shown to children"
And you can say the government SHOULD do that, but the government is so split that it never gets anything done.
AI is coming, and I'm sure it's going to continually erode what's left of what we call a society.
"Massive Government Intervention" used to have a much less scary sounding name called "Public Policy" or "The Law" and it gave us all kinds of good shit, things like the highway system, the National Park system, an electrical grid, building codes and standards, a currency that is practically the world standard, on and on. We have tons of things that would not exist without lawmakers who were well advised by experts coming together and putting pen to paper. Was it always good? No, nothing human made ever is. But their track record beats Silicon Valley's output by MILES.
> AI is coming, and I'm sure it's going to continually erode what's left of what we call a society.
Yeah and the frustrating thing is everyone working on it seems to believe this to one degree or another, and yet they're all still building it.
If you think a technology is fundamentally corrosive to society, then STOP BUILDING IT.
If you think your product will further the divides between people for profit, STOP BUILDING IT.
Like YOU CAN THINK. You can choose what to do based on what you believe will lead to a better tomorrow. Maybe if you did more of that and less soullessly chasing whatever some rich asshole wants you to build for a bag of money you'd be both A) more personally fulfilled and B) not furthering the collapse of civilization in your children's lifetimes you absolute cretins.
YES!
YES. Holy shit yes. We don't allow medicines or food additives that are substantially different from previous ones into our food supply FOR THIS EXACT REASON.
You know what would be even better than a ban? If these tech companies gave even the slightest ghost of a shit and maybe didn't immediately sprint to market with half-finished products the effects of which are completely unknown simply because there is no law that is explicitly preventing them from doing so and getting some goddamn information first about them and then bringing them to market. And that would be even better for this, what with a whole bunch of these companies also funneling money into think tanks dedicated to making sure the AI's they are burning money to build don't fucking kill us.
Holy. Actual. Shit. Like imagine being this incredulous at the notion of taking some fucking time to research the effects short term and long term of a new kind of entertainment BEFORE we just shotgun it out into the fucking world to our most impressionable and vulnerable population. Heaven forbid a modicum of reasonable caution, we have INNOVATION TO DO!!!
Christ.
We allow all sorts of food additives that are only now being banned because they turned out to be bad. I'm not saying we should allow them, but in general we do.
All of your complaints may be valid but you are living in some sort of fantasy land if you think that is going to happen. That's just not the way the world works.
Yeah I'm sure there was a legion of corporate whining when they said the same thing about banning children working in coal mines. We'll adjust.
> It's also the government insisting that they can know whats good for you better than you can.
I mean most AI researchers seem at best unsure about the technology they're building. That being said, I don't know if AI is bad for us long term, but I know damn well AI driven by profits is terrible for us right now and I'm cool with banning that for eternity.
> All of your complaints may be valid but you are living in some sort of fantasy land if you think that is going to happen. That's just not the way the world works.
Just because you can't imagine a better world doesn't mean everyone else can't.
There were content farms pumping out "finger family" videos where the same video would be uploaded thousands of times but with a different JPEG swapped in so this video is "Spider-Man" and this video is "Joker" etc. And kids gobbled them up apparently. They'd sit there watching finger family videos on autoplay for hours.
I just see a lot of people in this thread (and other related threads) seemingly suggesting that since these videos aren't already better than Pixar or some other arbitrary level of quality they've decided that the technology "isn't ready". I'm suggesting that is a short-sighted viewpoint. Kids (and that actually includes pre-teens and teens) are not going to care. If people generate funny or captivating stuff using this technology then they will definitely overlook the uncanny valley. And the volume at which this content will be created, re-mixed and re-shared will soon outpace any other kind of content source.
I find it a bit amusing since I bet a bunch of nay-sayers probably grew up watching equivalents of Rooster Teeth Halo Red vs. Blue, which was garbage quality for its time. I predict that we'll see at least one massive new studio rise using these "not ready" technologies in the next couple of years.
edit: (As another data point, consider the huge success of Roblox with kids. It certainly isn't due to the visual fidelity of the game. Consider the Roblox equivalent of YouTube that is going to exist pretty soon.)
I'm new to HN, but can see that the thread is no longer active (as in I can't reply). There was lots of great (and entertaining) feedback on the thread (especially about the terrible font). I would love to reply to those who fed back on the game and the page, now that I have ironed out some bugs with the game, and importantly changed the font to something that is much more readable (and with much better CSS). I've also released a mobile-friendly version of the game. Do you know of any way I can engage with those that commented?
Again, thanks for your interest and the two days of excitement I had looking at those analytics.
That's funny. I checked out your Twitter after the submission started trending on HN and saw that you posted the traffic stats. I actually wondered whether you would figure out where it was coming from, so I'm glad you eventually did!
Regarding your question I don't think there's any good way to engage with people after the thread has been "archived", especially since there are no DMs on HN.
Glad I could contribute to some excitement nonetheless. I posted it because I thought it was a cool and novel approach to a classic problem. I also really liked the article (and your homepage/games in general). Keep up the good work!
BTW - There's a mobile version of the game too now https://nonzerosum.games/tetrings_mobile.html
It was really cool seeing all those comments about the terrible font, and the game of course.
Unless you think you think you’re positioned to raise VC money in a down market, you’d only want to rely on something like with the plan of it being a low-investment “pop up” business where you make some cash now and shutter it when the model breaks.
Be careful!
Video will have a much much higher bar to pass than text and I think it will take some time still.
I am not so sure about that. The real hurdle at this point seems to be getting something like Dall-E to produce multiple angles and poses of a scene while maintaining characteristics of the elements in the scene. At that point it is an interpolation problem. I admit I am not an expert on these, but these do not seem insurmountable. I would not be surprised if we see something that can make very convincing short videos in less than a year. At that point, since a "short video" is really a scene, we are basically at film stage.
But silent films, no? Or where are things at as far as generated audio that syncs to the video? (sound effects, dialogue, music - each of which will have its own challenges)
Conceptually -- if you're generating the video from a script, you can feed the script and video into a foley and voice generator that recognizes "door slammed" and inserts the kind of sound associated with a door slam and that recognizes face changes and generates speech audio keyed to match. Those tasks will be solved independently on the way to long-form video itself getting sorted because you can exercise them with non-AI video along the way.
I'd be shocked if somebody hasn't already made some good demos of the foley and lip sync stuff.
This can be done with img2img and Controlnet on Stable Diffusion. But still a lot of manual work involved at least in my setup.
I doubt AI (by itself) will be generating movies like Disney movies any time ever, tbh.
But I don't doubt that it'll generate enough garbage appealing to enough people that it will account for Disney-like watch time.
Right now, the biggest issue with AI generation seems to be lacking the ability to generate a visual consistency with backgrounds, character faces, etc.
Without the consistency, it's really hard to a movie with any sort of narrative.
Without that, you're stuck with these small clips of 4 seconds.
I've seen some resort to face swapping for post processing, but it's very tedious at the moment.
Considering how many cuts are in modern movies, 4 seconds might be all we need.
Modern cinema cuts every 3-4 seconds. String a bunch of those together and you have anything from a TikTok short to a full-length movie.
https://www.youtube.com/playlist?list=PLUYhiBehAnloAfgA4Nuvk...
- https://www.twitch.tv/watchmeforever (ex Nothing Forever)
- https://www.twitch.tv/raycreationstv (Space Trek Show)
- https://www.twitch.tv/gptv_1 (90s TV Channel)
It would be nice to have that as a category on your site.
I found it slightly annoying that navigating back didn't work as smoothly as on YouTube, for example:
When the video is playing in the overlay, the URL does change when you attempt to go back, but the video continues playing. I'm not sure if that's intentional or something that could be improved.
https://www.youtube.com/playlist?list=PLUYhiBehAnloAfgA4Nuvk...
You can produce your own hyper-customized content on-demand and communities can form around sharing that content.
From video games, to text based adventures, to porn - I'd be surprised if humans are going to be competitive with generative AI for much longer.
Personally, I enjoyed going back and playing the old Pokemon games for nostalgia. Generative AI has completely displaced that genre for me. I can go on text-based adventures set in the world of Pokemon that have higher fidelity and degrees of freedom than any video game experience can provide.
The prompt:
> You are going to assume the role of a text based adventure game. The game is pokemon yellow. You will narrate the game and events and ask me for input to control the game. The gameboy was just turned on, go.
A snippet from the adventure:
(full adventure here: https://chat.openai.com/share/df0a1bef-229d-4f88-b25b-6e6af8...)
> Although Professor Oak advised you to stay away from Mewtwo, you can't help but feel a sense of responsibility and curiosity about the enigmatic Pokémon. You decide to return to the cave and try to communicate with Mewtwo to understand its intentions.
> Carefully making your way back to the chamber where you previously observed Mewtwo, you approach it with caution. As you get closer, Mewtwo seems to sense your presence and turns to face you. Its piercing eyes seem to stare straight into your soul.
> You gather your courage and attempt to communicate with Mewtwo, hoping to understand why it's agitated and if there's any way to help. To your surprise, Mewtwo appears to be receptive to your efforts. Through a combination of gestures, body language, and psychic communication, you begin to establish a connection.
> You learn that Mewtwo is indeed searching for something – a purpose and a place to belong. It feels isolated and misunderstood, and its immense power has brought it nothing but fear and mistrust from humans and Pokémon alike.
> As you converse, it becomes clear that Mewtwo is not inherently malicious; it simply wants to find its place in the world. You offer your support and friendship, hoping that this connection will help Mewtwo find the understanding and peace it seeks.
...
> You decide to invite Mewtwo to join your team and travel with you, hoping that together, you can help it find its purpose and place in the world. You extend the invitation, unsure of how Mewtwo will react.
> To your surprise and relief, Mewtwo seems to consider your offer seriously. It studies you intently, as if trying to gauge your sincerity and trustworthiness. After a moment of tense silence, Mewtwo nods in agreement, accepting your offer to join your team.
> With Mewtwo as a new member of your Pokémon team, you feel a mixture of excitement and responsibility. You know that having such a powerful Pokémon by your side could be both a blessing and a challenge, but you're determined to help Mewtwo find the understanding and acceptance it seeks.
This is right after battling Misty, I go straight to Mewtwo, team up, and then steam roll the rest of the game and go on an epic quest of self discovery with Mewtwo.
We find a bunch of ancient temples that are filled with manuscripts that talk about an ancient lost region. We go there and find it's the origin of all Pokemon, cut off from the world of humans and guarded by an ancient and mythical Pokemon Oriora.
The game was an incredible experience - far better than playing the original cartridges for me.
You're right, it does read like poorly written fan fiction.
Like I said, humans are currently competitive with generative AI.
I suspect they won't be for much longer.
> I'm not really sure how this translates to the death of the entertainment industry.
Implicit in my comment is the assumption that AI will continue to improve from here.
There is a clear curve forming in what generative AI is capable of.
For me to be correct, the curve holds for long enough that AI becomes a compelling story teller given user guidance/input along with corresponding stable generation of assets.
Given the poor fan fiction above is substantially better than where we were 5 years ago with generative AI, I think it's a pretty safe bet to say if we keep following that curve it'll become a pretty good dungeon master that can lead a quest in a meaningful number of worlds.
I don’t know if officially sanctioned self-insert stories would kill the entertainment industry, but they’ll certainly change how things work. Video games are already giving movies a pretty good run for their money, and that’s with an arbitrary skill threshold to keep people out…
http://localroger.com/k5host/casodycs.html
When you can have anything you want, the only thing you want is something you can't have.
Agreed that variable rate reward systems and competitive games aren't going anywhere. The sports and casinos are here to stay.
Pokemon wasn't about the grind for me, or the variable rate reward mechanics, I played it for the world building and story telling attached to childhood nostalgia.
I don't know you.
I don't interact with you.
I have no stake in what entertainment you consume.
Your decision does not impact me.
I feel no spite when you make these decisions.
Where's the struggle to overcome?
Where's the consequence for freeing the strongest, most manipulative, psychic creature in the universe despite warnings from the experts who created/studied the damn thing? He could crush your mind in an instant, but you're special enough that he instead submits without a fight? Isn't he supposed to be dummy-thicc Alpha or something?
Every choice you make is the Right one. You always win and nothing ever really goes wrong for you.
This isn't a game, it's an ego-stroking experience.
Or maybe it could handle DnD-style where it asks for checks with actual random dice rolls thrown in occasionally. Trying this with ChatGPT (it needed to be told to require checks multiple times) almost worked, but it still really leans toward success. E.g. attacking a goblin camp with the worst tactics imaginable did allow for "failure", but not actual failure. Decide to dance aggressively towards them and roll a 1 on performance? They laugh instead of filling you full of arrows. It's pretty funny but definitely removes all tension.