surely there is a limit to the detail. I mean, correct me if I’m wrong, but the objects all still have to be somewhat modelled first, so, the limit of detail on a specific object(regardless of how much else there is) is dependant on the modelling detail, just like with polygons?
I don’t think that they have to get smaller than the human eye can see unless it’s necessary for the story, in which case, they’d just model what is necessary.
That object has may to much charm and not nearly enough strangeness.
Well, the designs in that video weren’t all that great. But try and make scenery, like a beach, and you could have some outstanding realism. I really hope these graphics come to common games before I die. We can only hope 
Typical average-looking game graphics looked like this when I was born in '89.
It’s been 21 years & games have already gotten as far as to the uncanny valley. Give it 20 more & hopefully we should be out of it by then. Is it a stretch to assume you’ll live that long?
that was… ORGASMIC!!!
This is true. I guess “unlimited” in this case means that there is theoretically no limit to the amount of detail that a modeler wants to put into an object… it’s up to him when he decides to stop adding points.
But about the memory required to store all those points if all the modelers start to go insane…
Yeah I’m curious how this is meant to be “unlimited”. Surely all of those points have the potential to be moved, even if they only concentrate on the visible areas for rendering, don’t they have to include the whole “world” in update calculations? Gravity, erosion, object destruction, etc would still have to be calculated in real time whether they were displayed or not. Plus you’d have to store texture maps in memory for every single one of those objects. How much must that take up?!
^ Blurey & DvD players work on the principle of just changing the data that’s moving and rendering that part while leaving the other part of the picture normal.
That being said, the tech will work similarly; I.E. The unit is moving but the rest of the simulated environment isn’t. It can use the information from the last frame in the next frame.
The list goes on.
You could have a system where you model as much detail as you want, then tell the engine or whatever that this is a curve and have it fill in the rest. But that is my guess, however depending on context and your definition of detail you could still lose it as you zoom into a featureless curve.
Alternatively some kind of procedural generation of detail.
Something like vector graphics, IMHO.
I wonder if it’s even possible to translate vector graphics to 3D models. Bet it would be a bitch to keep textures from not glitching up.
There already is such a thing, this is also if I had to guess a similar system on which this technology is based.
They’re called splines. They’re lines, just like in a program like illustrator. The geometry assigned to them works on what are called iterations. The first level assigns 4 quads to the curve. Each iteration level exponentially increases those 4 quads (4, 16, ect) and there is no limit. This is just never done very far, because you get up passed 5-6 million polys and you’ll break just about anything.
As I said, my guess it that Unlimited Detail takes these free space points, draws splines between them, and increases the detail according to the resolution of the screen.
So, this OnLive thing seems pretty beast, I only have one thing to say, GTA4 maxed out playing flawlessly, sounds good to me.
Enjoy your input lag then.
unlimited detail:
minecraft:unlimited world map
(not really, after 30million blocks in one direction it gets very laggy.)
well when u think about it no u wouldn’t cuz if u look at any 3D object, lets say a tree, you see only the one side, if you move to the other side, u still only see that one side, you only see the shape as you move around, therefore, the texture changes will give the 3D impression… thats how i understand it
This is interesting: https://www.euclideon.com/press.html
The press release on this company’s new official website says they’ve secured the funding to create the SDK and expect to have downloadable demos out within the next 12 - 16 months.
Also, according to a comment that they made on one of their youtube videos, the stuff they showed was run on a single core laptop with no special graphics capabilities. I’m also assuming that means that they didn’t consume the entire hard drive storing all those data points. That pretty much convinces me on the static stuff. So the only real problem I have yet to see solved with this would be how it handles anything that moves. I saw a video that they made of a hummingbird but nothing involving any real physics calculations. If they can figure out how to make that run fast and smoothly, I’m investing my entire life’s savings into this.
Actually more like 9 - 13 months, since that press release came from mid-September