I'm just confused - what exactly are you aiming for? Things like advanced shaders? (blur, depht of field etc.)

These will all be possible with a material editor I plan on making. The material pipeline will not be fixed - you will be able to make crazy materials with shader graphs.

Check out this video of UDK's material editor. http://www.youtube.com/watch?v=GP0jX5nUB50 It is almost exactly what I'm going for.

The Xreal renderer is (was) updated constantly with new features but they get broken quite often and fixed rarely. If I am not mistaken, Coroner is wishing for a stable release with fixed goals.

Oh, I misunderstood then, that does sound better than my understanding of the F2P model.

What about the level editor, I hope you're planning on implementing some sort of BSP building brushes, that would really help with close quarters map design. :)


I'll try and be available for a chat.

I'm asking - what does Coroner want from the engine itself aside from technical stability?

I believe he just actually wants it to be done. It's a lot of work - sometimes an unreasonable amount.

I think TC(CQB) could be upgraded with some improved visuals quite easily if there would be someone who improves the ET engine in small steps with stable builds and limited goals.
The aim of such a project would be to easily recycle existing content and to leave the classic asset and map production pipeline intact. That might be not an optimal solution, but it ia a solution that would offer a real playable game ASAP.

Here's a list of things I would like to see in order of importance:

step 1:
*Radiosity light maps (known from HL2 and UrT Bumpy engine)
*Engine loads specular and normal textures if existing
*Specular and Normal map dependent lighting effects from radiosity light map and light grid.
(All this could be even ported and improved from UrT iOquake bumpy engine)
*improved screen based exposure (I can provide my code in Q3 src)
*improved CQB fake shadow mapping (I can provide my code in ET src)

(So, all the ingredients would be even existing)

step 2:
*Blur
*z-feather (aka depth-based alpha blending or soften smoke edges)
*Ability to load better skeletal model format, e.g., UT3/UDK psa psk.

step 3:
*Combination of radiosity lightmaps for bounced light with projection shadows
*UDK like solution with dominant directional (sun) light
(*alternatively COD4+ solution with distance based blending of directional lighting and lightmaps)
*VBO for better performance

Thats it.
Already with step 1, CQB could reach another visual level. It "just" needs a dedicated programmer willing to follow reachable (unspectacular) goals with stability and compatibility in mind.


Problems:
*Tactical shooters based on an a GPL engine are probably unplayable in public because of hacking (first hit decides most and wall piercing).

*e.g. UDK has much nicer and modern production pipeline and offers potential contributors to learn skills that look good in a CV.

The issue with continuing with ET is that if it turns into a serious commercial opportunity, we're out of luck due to licensing terms.

I've been looking at other various rendering pipelines that would be suitable. There are only two which have not escaped me. The first would be Deferred Lighting (a type of Deferred Shading) and the other would be 100% forward rendered megatextures (like from Rage.)

Megatextures are totally genius, and surprisingly simple. All textures are streamed in and out of memory onto a few texture pages that are typically 4096 squares. This actually has the added benefit of saving memory in some cases, as well as being computationally cheaper. Rage is 60fps on consoles, where as the COD games (et al) are 30fps. PCs rock consoles any day, too.

Megatextures are not considered to be totally pragmatic in the gaming world because consoles do not have the memory capacity to deal with it well, nor the ability to stream unmipped textures when necessary. If I do make it, though, it'll possibly be the only chance that people will have to work with megatextures because of Bethsoft's acquisition of Id. This is a huge opportunity that many people will not refuse. It's also easier to make really sexy and original scenes with such tools.

Check out the video for it: http://www.youtube.com/watch?v=b4ieFw2s7Fw

This isn't an unreasonable goal. It's a long shot, but it's worth it.

The megatexture idea sounds quite ambitious. I've played Rage and can say that it was really nice looking game but I somehow doubt that similar engine would be suitable for close quarter type of game. Map geometry looked fabulous at distance, but not that great when you came nearer.
Also I doubt that one man army programmer can produce something similar, at least not over summer.

Well, I think there a would be 10 goals 10 times faster to achieve and 10 times more important than megatexture.

I couldnt even see all that potential in the final product RAGE, though I put high hopes in it before. Truth is RAGE like other titles after idtech3 suck.

I can understand from an engine developer point of view that integrating old-school render pipelines in an even older engine doesnt sound appealing.

I don't think Rage sucked. The reason why the textures were blurry up close is because they wanted an even experience across consoles. You can download high-res mods that make everything look SUPER sexy.

Also, I'm not really doing 10 things. I've already got a physics engine in. That's a big one. The megetextures system is truthfully a lot simpler than people think. You're really just baking the diffuse, normal, and specular maps down with the lightmap. This technology has been around since Quake 1.

The only things different from it are the tools. You have normal static meshes that you can stamp things on, and then you also have tilemaps which you can paint and stamp on. The tilemaps are continuous geometry that have no seams. You use these tilemaps to hide seams on the meshes.

The tiles blend between each with a simple lerp. This is Q3-style terrain blending (anyone remember decidia?) The stamps are just kept around as dinky old-school decals until you bake it. Truthfully, there are only a handful of real issues I can think that I'll run into down the road.

You guys see impossibility. Why is that? Why not see opportunity? By entertaining me on this, you have much to gain, and little to lose.

I hope you will manage to get it working.

Yeah, it will be cool the rendering of it isn't very hard. The hard part is making the tools.

Affine, I don't want to discourage you. Probably I am just looking from another point of view. For me making the game is important, which even while the renderer is in the making or is getting updates, requires a stable toolset. That's what is all there for ET/UDK/etc and many ppl know how to use it.

I am looking forward to the possibilities with your approach, but in the case of megatextures not even tools are existing.

Nowadays most map geometry is made using 3D modelling applications like blender, maya and 3ds max, so this shouldn't be that big of a problem (unfortunately it takes awfully more time to skin geometry in modelling app than in radiant).

I am realizing that the tools will not be incredibly hard to make it's essentially radiant + a terrain editor. Stamps are applied to geometry with a simple ray cast (going from screenspace to UV space.) This is easy to do, and I can already do it. 3D painting is neither innovative nor nontrivial.

First I'll start out with just decal stamping (with no terrain.) This will at least give us somewhere to start, and will take far less time than including terrain.

Coroner, I'll make my own editor. Qt is a great toolkit