RDR2 with ultra-realistic graphics Gen-3 video to video Runway ML Artificial intelligence 

prompt looks amazing.... week after next ill have time to work on a photoreal prompt this good. Looks great bro I know the time it takes to make these keep up the good work this will pay off just keep at it.... looks great!!

Can you make a video on how you make these type of videos please?

@@jjj515 yea im working on the right prompts then ill help everyone I found some really insane settings now i have to text em but check my page you'll see some

Maybe next time you could do some cutscenes with Aurther? I'm sure we'd all love to see how he'd look!

how do you make it

Idk how young your parents or what ever are, but my one grew up in a time where they had to take the same bath for the whole family just about 20 years after world war 2, lived in a country that was devided into two, lived the reunion, saw the technology shift of 50's cars up till now with they own eyes, saw TV's becoming affordable and in color coming up, saw the internet coming up with all the Computers at home in general and still are being part of the current world.

@@MrGTAmodsgerman So your parents were born sometime around the 1950s in East or West Germany?

Were basically gods for them now

Runway has no actual reference of game geometry. It's making results based on existing reference. otherwise it couldn't generate a video with such fidelity of its own.

@@burkanx5546 50s yes, but no, later went to Germany.

I think It's because he specifies in his prompt that it is a western… So the AI apply some weird Hollywood filter from old movies.

Real

Oooh now that would be interesting!

Revision since it has more props and general map detail as a base reference for the AI. But yes, that would be amazing!

@deus_nsf good idea, that was my main worry is the environments might be too bare to work with

!Remind me in 5 years 😀

Hahahah “for a long time” thats the same thing that people said a year ago about Video Generators in general and here we are a year later

@@rombofn It takes a lot of time and power in kilowatts to process 15 seconds of video from a game. Do you want 60 times a second to process each frame at 500 watts? That's not gonna happen for a very long time. Certainly not on the PS6 generation, and not even on PS7

@@MrEleman When GPT4 was launch, it needed 40x more compute than today model (with same quality). So no, at this rate in 2-3 years from now this model quality will be in real-time. But the best model in 2-3 years wont be in real-time, but it will be in 4-6 years. Etc.

@@MakerTributeFan This is about the text model. Changing video data in real time is a matter of many years.

Its not but it will be, this is the path we will take in gaming to have another golden age of gaming again, we have stagnated the past 5 years

​@@rombofn I mean, ray tracing is cool and all, but I still think Crysis 2007 looks good lol. So yeah, it's safe to say that the market definitely waited for consoles to catch up. Aaaaand they'll be left behind again 😂

Good thing am on PC ​@@deus_nsf

Yes because this is only Video to Video it doesn’t alter the physics and animations (yet)

This technology appeared only a few months ago. Just imagine how the graphics could improve in the next few years.

Why? Not needed. Especially on bodycam, that game looks photoreal already

Yeah this is truly next gen NOT the next gen bullshit that we have now

No, it pretty much created everything from scratch, taking very little from the original frame. Look at older games with this technology, like half life. They look as good.

@@Project1nol But the AI hallucinates tons of stuff. Lighthouses become stone towers, pipes on walls become screens, computer terminals become trash cans. No artistic control.

@@Project1nol Distant people turn into tree trunks... Signs completely vanish... Clouds stick to the screen... These are all things I'm seeing immediately

Maybe it will. Rn it's generating based on the final rendered image from the game. If it gets access to the depth channel, geometry channel etc. it could be more coherent. Something similar was done with GTA 5. The video's at 2 Minute Papers

Don't forget that AI takes existing data from others in training to achieve this. It looks like a colossal ripoff, and i blame people for hyping it up if it bothers them. It's like, you don't have to capitalize on and eat everything you're fed...

Haha. Yeah it’s going to create a lot of debate in the long run. An interesting compromise might be for developers to create low-detail ‘scaffolding’ and then they themselves train an AI to replace the rendering with something more bespoke and fitting with their vision. It would still be soulless to an extent but perhaps closer to the likes of DLSS (which imparts little to no artistic directions) versus something completely free-reign, and that gives different players of the same game an equivalent experience. Putting a more positive twist on the usual talk of replacing jobs, if done right maybe (hopefully) it could lead to better balance in studios rather than people having to kill themselves delivering a game like RDR2, which was infamous for its brutal development. Wild times ahead.

bro ffs, they probably dont care, they do it for money

@@FVMods A counter-point to this is that it doesn’t have to be stolen. It is possible to create libraries of material for an AI to consume where you trace everything going in, and I suspect this might be the route huge publishers take, similar to how they share game engines across studios.

I agree with you, I wouldn't use this for modern games with good art, however for retro gaming this could be amazing.

This technology appeared only a few months ago. Just imagine how the graphics could improve in the next few years.

@@Skvodo I ain't gotta imagine crap! Ai will do it for me! Someone should generate how good ai will look in a few years.