Hey, LL. How bout some updated lighting?

Over and over, we've been told: "We can't have realistic lighting. SL is too dymanic. Any 3D engine that shows realistic lighting is PRERENDERING the lights and shadows."

Dynamic lights and shadows CAN be done. The C4 engine is doing it.
Look (boldface mine):

Eric Lengyel to Speak at GDC 2006

October 5, 2005
Eric Lengyel, one of the founders of Terathon, will be speaking at the Game Developers Conference in San Jose, March 20–24, on the topic of Advanced Light and Shadow Culling Methods.

Abstract
In the new generation of gaming hardware, lighting and shadows in graphics engines have become much more dynamic. For many types of games, it is NO LONGER ACCEPTABLE TO PRECOMPUTE ANY KIND OF LIGHTING OR SHADOWING that is presented to the player. This session discusses SOLUTIONS to the problems that arise when implementing a graphics engine that supports a COMPLETELY DYNAMIC AND UNRESTRICTED LIGHTING ENVIRONMENT. Topics include data structures, algorithms, mathematics, and optimizations that can be applied to a generic scene graph to efficiently determine vital information such as the set of light sources affecting the visible scene and the set of objects that cast shadows into visible regions. The methods presented are independent of shading systems and shadow rendering techniques.

Takeaway
Attendees of this session will leave with the knowledge of a plethora of techniques that can be IMMEDIATELY APPLIED in existing graphics engines.

Maybe someone from LL can attend this lecture.

Quick question, what method does There use for their lighting?

Game Developers Conference? Does this refer to the woefully slow PC gaming, or the lightning-fast console gaming of the nextgen systems?

You can wager that the PC games will get the short end of the graphics stick and we probably won't get great lighting unless SL goes 64-bit Windows only. Mark my words with graphite.

EDIT: OK, I meant 64-bit only, not Windows only.

What exactly does that mean? "windows only" Im sitting on a 64bit Ubuntu linux system right now. Although I do have to use a 32bit chroot for things like "macromedia flash", thats just because the developers are inhuman and dont give a furries @$$.

Hey Paulie, you offerin' to buy everyone a PCI-X video card with the latest PS/VS versions? That'd be damn spiffy of you. 'Cause I don't think you're gonna be pipin' the bandwidth needed for that through an Intel Extreme or old rusty Geforce 4, which is probably what most of the world plays on.

You can yell about research papers until your blue in the face. Try implementing an idea like this on a cross-platform *MMO* (There are many different things that can be precalculated in respect to light when you have ALL of the data in a mostly guarenteed "fast" access medium like, say a harddrive versus a network) that has to support many different types of machines.

Get back to me with your results. Otherwise, please, stop acting like cold fusion is obvious to the casual observer, 'k?

Sorry just had to post again, The Reality Engine does all of the lighting in realtime, hdr etc. And works perfectly fine on my Nvidia 5200, Just the more advanced shaders are not compatable.

1.7 contains a Full Bright feature that lights a prim per face (texture). It has the effect of shining a floodlight on the face so that's it's lit without polluting the surrounding landscape with glowing light rays.

It works great for windows, trees, vendors et cetera. If we can educate people to it's existance, we might severely reduce the light pollution on The Grid All Hail The Central Grid.

Granted, lighting still needs to be fixed. But it's been relegated to the mythos along with improved group features, havok2, an open communication protocol, web on a prim, text on a prim, local storage, mono, damn this list is getting long. Well at least the programmers are having fun in the challenging and innovative atmosphere that LL fosters. I wonder how community relations people list items such as "oversaw the successful two year program of convincing users that foo would was a good idea, being developed, and would soon be implemented" on their vita?

Fullbrights are a hack. Fullbright textures are simply ignored by the lighting engine. No lighting calculations are done for a fullbright surface. Hence the framerate improvement.

Fullbrights are a step *backward*, away from fully dynamic lighting. Fullbrights were cute in Quake 1, but not in a modern 3D engine.

Fullbrights are part of an old lighting model, where selected surfaces are lit using one technique and others are lit using a completely different technique. This lighting model is a relic from the early days of 3D-accelerated game engines, and can never achieve the realism we need.

Fullbrights are not compatible with modern lighting and shadowing features, such as HDR (high dynamic range) lighting and parallax (virtual displacement) mapping.

Cory has claimed that we will be moving towards a "materials-based" texturing system in the future, where surfaces will be covered with multiple texture maps, such as gloss maps, reflection maps, displacement maps. Fullbrights completely waste the advantages of a materials-based texture system.

We should be moving towards fully dynamic, per-pixel, scene-based lighting and shadowing, not backwards towards Quake 1.

LL should add a pre-processing button.

Basically, you can click on it and it will use your client to go and calculate all the lighting textures which you can then statically apply afterwards.

Or, the 'textures' could be procedural in that they have 'lighting information'. But in a way that is low bandwidth and low storage costs, but it is all statically compiled.

True, you won't have dynamic lighting, but as Aimee and many others have shown, dynamic lighting isn't necessary to give something proper mood.

I'd like to give this a spin, it's like FREEZE in audio programs.

http://www.sweetwater.com/expert-center/glossary/t--Freeze

To be honest, although I like this idea, I'd rather see LL use it's development power on building/adapting a modern rendering engine, instead of adding more and more hacks to the old one. (This is assuming the renderer 2.0 now in development actually provides techniques like dynamic lighting, specular/bump mapping, vertex and pixel processors etc... and not just a hugely increased viewing distance.)

Highly, highly, HIGHLY skeptical here.

Quite frankly, this is the kind of thing where you sit back, and wait-and-see, does this technique work with 100 lights simultaneously? 1000 lights? What performance hits do you see? Are there other issues, like which enviroments can handle this technique?

Does this technique work on 'Integrated Video' -- last I checked, a huge number of machines were being produced with dodgy onboard chipsets; so you should factor around them existing.

You do not rush in and code a solution around an experimental technique, likewise 'fullbright' or prelit vertices are not a 'hack', they have perfectly valid uses. (Especially on surfaces which are supposed to be light emitters -- plus do you want every bright surface to emit light? Look what mess that has made in SL so far)

-Adam