Stuff to make a HD tremulous

[jm82792]

Well this is the issue.
A new rendering engine is nice, I understand that to a vast extent.
For fun I make CG scenes(right now it involves brute force of 10 million polygons for an ocean scene) and try to get them to look realistic.
Anyways I understand on how bump mapping/normals mapping would really help along with displacement mapping combined with an animated texture, stuff like bullet holes and such.

But just like directly contradicting an engineer's philosophy(in some ways, sorta..) we can have a higher polycount and get great FPS with machines that have a $50+ GPU. Efficient? No. Works well within reason? Yes to a decent to good degree.

Frankly, if you want a higher quality render output, it's not so much the models that need to be changed. Rendering lots of polys is still hard and a waste of time. The client engine really needs to be written to include some form of modern bump mapping such as parallax or normal mapping so that more complicated geometry can be simulated without significant increases on hardware requirements. Displacement mapping would be really nice. Solving graphics ugliness by increasing game model detail is the brute force method and isn't really as effective as other, cheaper (in terms of hardware load) solutions.

Also, none of these improvements are going to look really good unless and until we get dynamic lighting and soft shadows.

'High Definition' is marketing bullshit. As has been noted already, its only real technical meaning has to do with display resolution. When used as a descriptor in any other context it's meaningless (do commercials for 'HD audio' bother anyone else?).

[Pazuzu]

But just like directly contradicting an engineer's philosophy(in some ways, sorta..) we can have a higher polycount and get great FPS with machines that have a $50+ GPU. Efficient? No. Works well within reason? Yes to a decent to good degree.

So what you're saying is you want higher-polycount models, and that this has nothing to do with a "HD tremulous".

[CATAHA]

under HD tremulous you mean just bigger and more detailed textures or what? If you want make 'better' trem with improved visual style and animations 1st remake client side and implement new rendering engine, for example xreal.

[Pazuzu]

under HD tremulous you mean just bigger and more detailed textures or what? If you want make 'better' trem with improved visual style and animations 1st remake client side and implement new rendering engine, for example xreal.

So changing the rendering engine will magically make the textures more detailed? A new engine can only do so much, and while it would be nice, treating an XReal port like some sort of silver bullet will get you nowhere. Good engine + shitty textures = shitty graphics. Since modern game graphics place emphasis on a low polycount, almost all the detail is in the textures.

[CATAHA]

Im not saying new engine is all you need. I said you need it as 1st step! Good huge resolution textures on q3 engine - waste of engine power. If you talking about modern game graphics, then dont forget about modern shaders. They do a lot of work to improve resulting image. If you take hires textures (for example) from Q4 and put them in q3 engine you will never get even half of they 'quality', caus no 'bumping/specular/etc stages implemented. So 1st - upgrade engine, 2nd - improve textures/shaders accordingly to new engine and only in 3rd talk about 'ultimate HD version' of tremulous.

[Pazuzu]

So 1st - upgrade engine, 2nd - improve textures/shaders accordingly to new engine and only in 3rd talk about 'ultimate HD version' of tremulous.

Surprisingly I agree with this. +1 for you.

[jm82792]

I apologize yet again for being in a tired daze,
thankfully for myself(Like you care lol ) the slave labor will end soon.


If we port trem to a new engine(Xreal has lots of bugs from what I hear, bad idea?) we won't have to worry about new textures because people will gain an interest in helping trem and see the ugly bottleneck of crap textures.
Plus materials, trem has the capacity but they are a pain in the butt to utilize and create.
Materials will make the game look really nice if we have more variety and control.
Now a new engine gives us new rendering capacity, in speed because we won't have a decade old engine and
features like bump mapping,(insert here, I could go on forever) ease of use, etc. Plus maybe a less confusing (at least I think) engine.
Porting as everybody knows will be a pain,
and we need a competent leader who coordinates getting trem to the new engine.
Do NOT care about how good it looks initially,
but what it can look like with new textures, models, etc when you have a more capable engine.
If we had the same old trem in a new engine we'd be in good shape because we can bug test, develop,etc and make it more visually appealing.
(I know nothing about porting, I do know 3D and how it works)

Keep in mind bloom, and such look like crap when a newbie uses all the idiot proof features that look bad unless used properly.


Now realistically within the reach or capacity of a single individual(yeah a single guy can port but that's not going to happen) we can create/acquire textures(they have to be 100% free, proper licensing or made yourself.  ) and create a patch/mod/upgrade with a slew of new textures to replace all the default map ones. Most custom maps utilize default textures so many maps will look better.

New models are more complex, not very, but if your going to upgrade their textures don't waste your time, redo the models then texture them.



The complainers can stay on pure servers and use their low res stuff,
better yet the devs will allow us to use "this" mod or, I doubt officially integrate it into trem.

[Tremulant]

we won't have to worry about new textures because people will gain an interest in helping trem and see the ugly bottleneck of crap textures.

Oh, that's ok then.

So, since xreal isn't overly mature, is a touch on the buggy side and has a few compatibility issues, what are you going to use instead?

[jm82792]

I have no clue.
There is a website that has been bug fixing and imrpoving ioq3 for awhile.
Nothing special(well it's special but nothing graphically), maybe they will get around to getting it to render better.

[Pazuzu]

There'll probably be something. Other than XReal, I can think of Syntensity and Cube/Sauerbraten off the top of my head. Sauerbraten is the best (IMO) because it's fast and (relatively) lightweight, but translating the maps... ugh. Still, I'm not a developer, what do I know.

[jm82792]

If we can get the mesh into it without it being some giant mess then texturing and doing materials should be easy.
Well unless the mapping software(whatever it is) has some quirk.

[Pazuzu]

If we can get the mesh into it without it being some giant mess then texturing and doing materials should be easy.
Well unless the mapping software(whatever it is) has some quirk.

That "quirk" is called octree mapping. Sauerbraten uses a completely different map format- maps are made completely of huge arrays of cubes, where the vertices can be pushed and pulled to make non-blocky shapes. The maps can be really high-quality stuff (check out Skycastle, that thing is EPIC), but it's a completely different format.

[jm82792]

Well we can make new maps, but the whole concept seems far out and not something I've ever heard of.
I do not see their method as any easier, what you see is what you get mapping seems rather crazy.

Soon grandma will own a quadcore, you'll own a 16 core CPU, and GPU rendering will be norm.
Anybody up for porting trem to the Blender Game Engine ?

[Pazuzu]

It seems kind of arcane at first, but this video should help. A bit. Maybe. In any case, transferring all the maps won't be pretty, but it's definitely doable (especially with the existing maps as blueprints).

[jm82792]

It seems like it's effective enough although a higher level of control can be done with plain old modeling.
According to Wikipedia the engine renders fast and from what I see it looks decent.
I know nothing of how to port trem over or quake coding though

[Pazuzu]

Whoa. I just realized Aardappel (who created Cube/Sauer) worked on Far Cry. He's got some inside knowledge.

[jm82792]

I'll help but I won't be driving anything.

[Tremulant]

Soon grandma will own a quadcore, you'll own a 16 core CPU, and GPU rendering will be norm.

Why this obsession with the future of computing? What matters is what's affordable and current, i don't care about your granny and her quadcore box.

[Pazuzu]

Soon grandma will own a quadcore, you'll own a 16 core CPU, and GPU rendering will be norm.

Why this obsession with the future of computing? What matters is what's affordable and current, i don't care about your granny and her quadcore box.

Admit it, you can't help but admire her l33tness.

[CreatureofHell]

Soon grandma will own a quadcore, you'll own a 16 core CPU, and GPU rendering will be norm.

Why this obsession with the future of computing? What matters is what's affordable and current, i don't care about your granny and her quadcore box.

Admit it, you can't help but admire her l33tness.

+1 

[jm82792]

No obsession. The reality is that it's not a bad idea to port trem to a demanding engine if in a year or two it's requirements won't be that much.

Besides my granny doesn't own a computer.
I am trying to attempt a form of sarcastic humor regarding that what we consider good this year is crap the next.
If you ported to the BGE as crazy as it sound,
current PCs can run it fine(very broad example) and in a year or two it will become even more sane sounding.

[CreatureofHell]

No obsession. The reality is that it's not a bad idea to port trem to a demanding engine if in a year or two it's requirements won't be that much.

What if 1.2 shows up in a year or two?

[jm82792]

1.2 is as we can currently judge a gameplay and graphical upgrade that involves manipulating the current assets,
and from what I can understand they are not touching the current engine's code.
Plus they(from what I understand) are not going to touch textures, upgrade models,
well minus the new weapons and the new human model.
None of this stuff is easy but it would be rewarding to port or more realistically upgrade the current assets,
if we upgrade the current assets we can squeeze the engine with some brute force.

[CreatureofHell]

1.2 is as we can currently judge a gameplay and graphical upgrade that involves manipulating the current assets,
and from what I can understand they are not touching the current engine's code.
Plus they(from what I understand) are not going to touch textures, upgrade models,
well minus the new weapons and the new human model.
None of this stuff is easy but it would be rewarding to port or more realistically upgrade the current assets,
if we upgrade the current assets we can squeeze the engine with some brute force.

My point is that, as the textures and models (for human weapons at least) are already getting upgrades in 1.2, if you do all these changes for 1.1 then it all seems a bit pointless.

[Thorn]

About the off comment that magically dumping Trem into XReaL won't improve texture detail:

IOQuake3 appears to use software mipmapping as respect to archaic hardware that wasn't compatible. This makes the textures appear 2 times as blurry as they actually are, especially at a distance. It's not really a big difference, and could probably be implemented into trem without much issue, but I felt it was worth mentioning anyway.

@Fox One

I disagree, shaders such as parallax/displacement have a far more significant impact on framerate, compared to brute-force polygons. The majority of Tremulous' theme is 'tech', that includes metal base matt textures, metallic trims and the like. These shaders weren't made with that purpose in mind, which is why they apply far better to games like Crysis and FarCry, where running shaders on masses of terrain is of course cheaper than ridiculously high polygon counts.

[freezway]

just wanted to say:
GET 1.2 OUT FIRST!

Then work on the engine.

[CATAHA]

Just wondering why you not talking about '1.2' and 'working on engine'. =)
I am as many other players want 1.2 release. I dont like 1.2, but when it will be released, we can work on mods and stuff to improve it. But its anyway sad to hear, that players just w8i'n 1.2 instead of some 'work'. It means they just want new release in any case, but noting more. =\

[jm82792]

Interesting..
IOQuake3 is the bottle neck for textures, although bigger ones can somewhat help.

About the off comment that magically dumping Trem into XReaL won't improve texture detail:

IOQuake3 appears to use software mipmapping as respect to archaic hardware that wasn't compatible. This makes the textures appear 2 times as blurry as they actually are, especially at a distance. It's not really a big difference, and could probably be implemented into trem without much issue, but I felt it was worth mentioning anyway.

@Fox One

I disagree, shaders such as parallax/displacement have a far more significant impact on framerate, compared to brute-force polygons. The majority of Tremulous' theme is 'tech', that includes metal base matt textures, metallic trims and the like. These shaders weren't made with that purpose in mind, which is why they apply far better to games like Crysis and FarCry, where running shaders on masses of terrain is of course cheaper than ridiculously high polygon counts.

[Thorn]

Not true,
It would most likely be trivial for someone experienced with GL to patch in hardware mipmapping (assuming such a patch doesn't already exist). Mipmapping isn't really something to change engine for, and it certainly does not inhibit or 'bottle neck' textures in any way.

[Odin]

How many times do I have to say this? A port to XreaL would not require new assets. It would look no different, but it would use code that is designed to run as much graphics on the GPU as possible, speeding up the game vastly. The enhanced material effects and special model formats can be utilized later.