Real Time GI idea for skyrim?

  • Author
  • Message
Offline
Posts: 24
Joined: 22 Oct 2015, 04:14

Re: Real Time GI idea for skyrim?

Yeah I do know how they draw them. I just don't know how this one game draws them, it sounds like a bad situation to me. Well I tried, if there's only screen space... Good luck!

Offline
Posts: 24
Joined: 22 Oct 2015, 04:14

Re: Real Time GI idea for skyrim?

(Also I don't mean like for rendering the object. I mean like getting an origin for said object)
I'm guessing you got that though

Offline
User avatar
*blah-blah-blah maniac*
Posts: 17427
Joined: 27 Dec 2011, 08:53
Location: Rather not to say

Re: Real Time GI idea for skyrim?

I can't get origin of object, it's too much work (you can't do this with just SKSE) and memory requirements which already problematic in the game.
_________________
i9-9900k, 64Gb RAM, RTX 3060 12Gb, Win7

Offline
Posts: 24
Joined: 22 Oct 2015, 04:14

what if?

If this was done on the special edition would it be any better for memory issues?
From what I know my idea would only take a bit of memory, though to be honest I don't know how the creation well enough to be for sure with this.
I would still like to explain my idea, though I know it may not help much it still could lead to another better idea.

(Very first idea which is processor intensive and still is probably good to look at)

My idea was to get (from the engine, or file of the game) the lod for the said object (not including dynamic objects)
and use the objects position(origin) for a length from other objects around the player, near by the screen, or behind the camera a ways.
Then find the sun and other near by light sources.

Now for this example I will be using two trees, tree a which is on screen and tree b which is off screen.

With the now known lod texture for the tree B, you can then project a texture (black and white (Ao),or with low res color) on to tree A.
of course I still need a direction which the lod texture is applied from.
I get that from Tree B the one I'm casting the texture from and the Tree A which I am painting on to.
So I add Tree's A origin and Tree's B origin and then normalize the sum and get the basic direction paint direction.
This paint is done for each pixel sadly making it even more costly, if doing a length check for near by off screen object wasn't.
scale the textures lightness based on the length from the light and length from Tree A to Tree B
and length from the other object it is bouncing off of.
Lastly do phong and basic diffuse. (for static objects this does not have to be updated per-frame)

p.s. the "lod texture" can be swapped out for a basic sphere.
This sphere will still needs to be colored for each object in the game but for AO can be used not in this way.

(probably another costly idea x2 again still probably good to look at)

So this will just go over ao, you can still get the same lod textures for hq stuff...
okay so I take a theta of the cameras 0,0 pixel and it's current viewing direction.
Then take the off screen object theta and the viewing direction and see the difference (subtract, add, or something to compare the difference).
After that, based off the difference apply some darkening shading (maybe like a phong) towards the on screen object.
(here is where you would apply the lod texture to make the phong a certain shape, or color, or even project the texture for better AO, or even GI though costly.
You could also do per-vertex shading for the ao, or the GI again which is much less costly than a projection.)
The darkness is based off of the theta difference and the off screen object length from the camera and other objects lengths that I'm am rendering.
The further away and object is I'm rendering, then the less the shading is on that object.
Then apply the lighting shadows and shading for the final image.

Offline
User avatar
*blah-blah-blah maniac*
Posts: 17427
Joined: 27 Dec 2011, 08:53
Location: Rather not to say

Re: Real Time GI idea for skyrim?

SkyrimSE will not have memory issues to do that, but it is obfuscated and i will not make any closer by features to old Skyrim, so why to bother if it will look like ass anyway?
About your ideas, if i understood. Far from all objects have center points and their shapes are very different, one can be in the camera, but actually somewhere far away mountain or flat surface of river. Okay, let's say in some prepass calculate their cosine transform distribution of visibility from every virtual side to calculate how much it visible for one point of space, how to get light distribution of it? Object can be much bigger than lights on it and can be in shadow. Or these objects may be crossed in direction of tracing by other objects and any simplification kills that crossing, they start to hide or reappear without any rules, making gi flicker. Presenting objects like set of some mathematical points do not work for gi computation and can't work, it's not gi, it's environment map then, which can be computed way simpler. Finally, this do not have any huge improvements of performance to usable levels. Instead of asking me to implement such things, you should try make simplest demos in any game engines or calculate everything on paper. If things would be so simple, then games already would have gi and raytracing. This is too much complicated problem, any approximations found, any tricks, still need huge calculation power not compatible with games. Modern cpu and gpu not designed for fast raytracing of non procedural objects and even with procedural like fractals it's very complex task. Any attempts to make gi or even simplest indirect lighting includes various optimizations which must be made only with sources of the game and they never applied globally, only in some places or very low resolution to not hit frame rate much. I have spent myself many months thinking how to make generic indirect lighiting at playable performance and all ideas still unrealistic to implement or quality not worth to bother. Please understand, it's too much hard task and within limits of some game which i didn't made it's no more than wasting of time. So many graphic developers in the world and nobody can find any suitable realtime gi solution and all known have so many precalculations and simplifications, if you check papers...
_________________
i9-9900k, 64Gb RAM, RTX 3060 12Gb, Win7

Offline
Posts: 24
Joined: 22 Oct 2015, 04:14

Preview I guess?:

Preview I guess?:

I will probably post my response tomorrow... I don't like the over all approach of defeat to me. things like "I will not make any closer by [adding] features to old Skyrim so why to bother if it will look like ass anyway?", and my response because it's a lazy way of looking at it when people are adding pbr to "[it] look[s] like ass" Marrowind with it's bad textures.

I guess a bit of this too, to show what is ahead.

Object can be much bigger than lights on it and can be in shadow.

(
If you meant that light size then, if the object is big and I know that the light source is small then I can just ignore it really, or somehow fake it or make a bright circle of pixels averaged in.

Also if you get the visibility of it can’t you take the distance and just multiply that in some way or scale to get the size of the object. Heck with that it might just be good to put a copy of each object in a room somewhere for references.(random thing I added)
)

also in the preview I guess is what it is? an end statement!

I have spent myself many months thinking how to make generic indirect lighting at playable performance and all ideas still unrealistic to implement or quality not worth to bother.

(^ if no one ever “bothers” then no one will ever get the task done, or the problem solved. The more iteration and the more ideas you take out of picture that wont work the closer you are to the answer. Another person making you think can only improve and take you down a road you had never seen before. Although you might have seen it before and walked it.)

Please understand, it's too much hard task and within limits of some game which i didn't made it's no more than wasting of time. So many graphic developers in the world and nobody can find any suitable realtime gi solution and all known have so many pre calculations and simplifications, if you check papers…

(^I have checked papers (real time and not real time) and have had many thoughts that are usually simple.
Though If we don’t stop looking and even look in the place we think we won't find it in a place that it just might be.)

(I completely understand the struggle it sucks and everyone want a solution, that’s why in the second one I mentioned AO instead of GI. To start on a simpler and similar problem so, that it could even be tested faster. I know once you have half of the problem done it might be close to being done.)


Thank's
Post Reply