pre-renedered or dynamic lighting?

which do you think is better, pre-renedered environments or ones with dynamic lighting like doom 3?

when you think of 'better' you must take in account of performance, how it looks, complexity of engine, everything! and remember this is a personal question take these things into account for yourself. i want to see what members think.

oh and specify all your answers please.

remeber this is also a question of how playable one is to the other as well as the editing aspects

---

Bullet Control: $5000 for a bullet.
"I would blow your f**king head off! ...if I could afford it. I'm gonna get another job, start saving some money... then you a dead man!"

Add an "I don't know what the hell is going on around here" option, and you've got another vote! :razz:

I have never mapped for a dynamic lighting system, and pre-rendered seems to work just fine to me...

---

I tried sniffing coke, but the ice cubes kept getting stuck in my nose.
http://www.dimebowl.com

i have to say that i cant decide. the perfect engine for me would be one that had pre-rendered environments (such as static objects, buildings landscape etc, and static props all pre-ren) but real time lighting for models and physical objects

---

Bullet Control: $5000 for a bullet.
"I would blow your f**king head off! ...if I could afford it. I'm gonna get another job, start saving some money... then you a dead man!"

d00d... where's my option? oh yeah, also, I agree. Pre-rendered static objects and dynamic on dynamic objects

---

I tried sniffing coke, but the ice cubes kept getting stuck in my nose.
http://www.dimebowl.com

Everything dynamic. It is the future of gaming. Static environments will be a thing of the past. In the year 2000.....................

Pre-rendered.

Given how fast Doom3 ran, it was either horrendously unoptimised or
technology is simply not ready yet. The level graphics themselves were
pretty horrible and it seemed that every single advancement that we've
had since Half-Life was thrown out the window for this single feature.

It's not like the lighting looked good either. The edges of shadows
were too sharp and the lighting tended to be either on 100% on a
surface or none-existant. You would never have a faded shadow.

Look at FarCry. The lighting and environments in that game were FAR
better than Doom 3s and it used pre-rendered lighting. Using tricks
such as projectors you can create some very believable lighting that
looks ten times as good as Doom 3s lighting and runs much faster. Sure,
there are circumstances that Dynamic lights would be fantastic, but for
the most part when wandering through D3 I was thinking to myself "I
could do this all with static lighting...".

A mix of static and dynamic lighting could work, but every single light
dynamic? That's a horrible allocation of resources, especially when you
want more than 2 lights in a room.

---

-KingNic

-Slapping polygons together incoherently since 2000

Well... I WANT Real Time Global Illumination and Caustics, Sub Surface Scattering, PerPixel Displacement, Light scattering volumetrics and HDRI.

But I'm not going to... so Baked lighting for me. You can always get more out of PreRendered. ALWAYS. Put the burden on the compile not the client, that's what I say.

You can always get more out of PreRendered.

Always? Ha. Where's a pre-rendered day/night cycle? Where's a pre-rendered on-the-fly color shift? Pre-rendered attachments to moving objects? Pre-rendered lights that are dynamically shifted by sounds and changes to global shaders and variables? You can do none of these things. Lightmaps have their strengths, but to speak in absolutes is going to get you nowhere :razz:

Given how fast Doom3 ran, it was either horrendously unoptimised or technology is simply not ready yet.

Having a min spec listed of a gf2-equivalent card (gf4 mx) wasn't the best direction, either. If I were to min-spec a game like D3, I'd personally say use a high-level gf5 or higher (which is higher than most titles, but certainly not yet a mark of 'not ready yet'). But even then, there's far more at work than just the lighting and shadowing. Surprisingly, there's some CPU bottlenecks as well. Texture memory is also a huge factor - When you turn d3 texture compression off, no video card on the market can handle the amount of memory you're throwing at the card. (well, ok, a few 512 cards are out now, but you get the idea). D3 is pushing immense power all over the place. Rendering is just a small part of that, and can easily handle hundreds of thousands of triangles on screen with plenty decent lighting.

The level graphics themselves were pretty horrible and it seemed that every single advancement that we've had since Half-Life was thrown out the window for this single feature.

Like...? It certainly looks terrible with all the content scaled back. But what game doesn't? I can see calling some of the lab stuff a little redundant, but to complain of end-game stuff (hell, dig site, etc.) is blasphemy! :smile:

It's not like the lighting looked good either. The edges of shadows were too sharp and the lighting tended to be either on 100% on a surface or none-existant. You would never have a faded shadow.

I didn't see it in Doom3, it's clearly not possible! :razz: I'd say this is half conscious direction, half min-spec. id wanted a very harsh style to everything, so there aren't many fill lights and, of course, when you're running a min-spec guideline on a card with the power of a gf2, you can't afford much lighting at all. But all it takes is a few fill and multi-directional lights and you can get much more subtle shadows.

Look at FarCry. The lighting and environments in that game were FAR better than Doom 3s and it used pre-rendered lighting.

Meh, I found Far Cry to look decent enough, but place it well behind D3 and HL2.

Using tricks such as projectors you can create some very believable lighting that looks ten times as good as Doom 3s lighting and runs much faster.

Every light in d3 uses a texture much like a projector system. Many of d3's best shots use a combination of these in addition to regular lighting. It'd be easy for a third-party developer to just swap out more light textures (since it's using textures anyway, it's 'free' to use these instead) for these, which also can help immensely toward the appearance of softer shadowing.

Sure, there are circumstances that Dynamic lights would be fantastic, but for the most part when wandering through D3 I was thinking to myself "I could do this all with static lighting...".

Some of it could, I'm sure, but it'd look nowhere near as good. And d3 was very good about using the technology they were given; you'd have a very hard time lighting d3 statically.

A mix of static and dynamic lighting could work, but every single light dynamic? That's a horrible allocation of resources, especially when you want more than 2 lights in a room.

Developing content for any game is nothing but tricks, and using a dynamic lighting system is no different. A huge hit comes from shadow volumes, so even a small thing such as unticking the "cast shadows" box gives you immense gains. As for two lights in a room, again, that was half by design and half min spec requirement. Make custom content for d3 and you can take things much further. Hell, I 'only' have a 9800 and I can easily run d3 at high quality 1024x768... That leaves a lot of room for graphical polish a gf3 or gf4mx can't even begin to handle.

---

www.button-masher.net

i personnally think that many of the models in D3 just didnt looks real enough (forgetting the fact that they were demons from hell). dynamic lighting in D3 wasnt done right. the main reason for that is i dont think that the average computer is not quite ready for so much graphical detail yet. i still think that a mix of the two would be the best. you would get the amazing look of real time lighting on dynamic models but the performance bonus for usually texture rich environments.

Prerendered has the advandage of taking off the work off the
client. With full dyanmic lighting, your pretty much doing a
light combile of what you see every frame. Not the most efficient
way to do it, and make large outdoor areas without a ton of fog, pretty
much impossible to do on the standard machine. I personally like
the idea of doing most of the lighting at combile. Granted, some
lights should be dynamic, but not them all.

Until Later...

---

Dark_Kilauea
DVS Administration
http://www.dvstudio-production.com/

Photon mapping all the way!

As of right now, I see shifting from pre-rendered lighting a good
thing. Most peoples video cards (sadly not mine) have enough fill
rate to give pretty high preformance even with dynamic lighting.
Even then you could probably include it as a switch to turn
on/off.

Most game engines are getting away from any pre-compiling at all.
Far Cry, Serious Sam 2, Doom 3, etc. With the right data
structures not only will your performance be better, you can do alot
more things than you can with quake style bsps. Why Valve is
still using them I have no clue. Most engines by now should be
using portals, and probably occlusion culling would be even
better/simpler to map for. Take Far Cry as an example. The
average level load takes maybe twice as long as HL2, but instead of
lasting 5 minutes, it takes maybe an hour to get through, and is just
plain a ton more expansive. Plus development is easier because
you can see the product very soon after just a 30 second level load,
compared to having to compile it and wait maybe 30 minutes if you are
lucky.

One of the problems with Doom 3's lighting was that it didn't even look
that good. Probably newer games will come out with better
lighting that is more realistic, realistic soft edged shadows, global
illumination - not just direct lighting. And probably photon
mapping is way too resource intensive, but with a clever
implementation, it'll eventually be the best way to go. Even
above and beyond realism, dynamic lighting gives much more flexibility
than static lights do.

Oh and by the way Far Cry has both dynamic and static lights. A
pretty good blend too since it can run well on high and low end systems
reasonably well.

http://web.telia.com/~u31225218/fc/dynamic_lights.html

---

To get something to work, sometimes you just have to beat your head against the wall longer; the skin grows back, but the brick doesn't.

Source hates soup!

It was an awesome experience to witness dynamic lighting working in FarCry for the first time. The swinging light fixture with flickering shadow is just a beauty to behold.

Half-Life 2 looks good too, but I think it'll look even better with dynamic lighting implemented.

---

"The greatest thing you'll ever learn is just to love and be loved in return." -- Toulouse-Lautre, Moulin Rouge

I just love the concept of dynamic lighting, so it gets my vote.

The more real - the better IMO.

---

www.senshiserve.tk

Dynamic light, radiosity, and everything else is the ultimate goal. From there, I think, the only place to go would be varying interaction (suits and such).

---

Blame it on Microsoft, God does.

Well technically you can always get more out of pre-rendered. Especially when the whole game is pre-rendered. Besides, you could have animated pre-rendered light sources for day/night cycles. Even a map in half-life 2 run in Mat_Fullbright has a wealth of pre-rendered lighting. Without pre-rendered lighting, all the walls would look like cartoons, the props would look rediculous, the sky would be a solid color, the grain of the wood would just be god awful. I would much rather have a map with a room lit with a top of the line renderer, and baked into a high resolution map than to have a couple of doom 3 lights.

Dynamic shadows are of course necessary since it's pretty much impossible to pre-render character shadows, but if the object aint moving, there is no reason to have dynamic shadows cast by it.

I'm sure in the nearer future what we'll see is the map having it's lighting baked and then as technology progresses better and better character lighting, but really there is no reason to dynamically light a table in 99% of all situations.

Animating a day/night cycle? You're mad! :razz:

You're already chewing up file size and memory when you're computing pre-calculated light maps. Now imagine the memory required to smoothly transition all the lighting, shadows, reflections, etc in the map for 24 hours of daylight. That's 3600 seconds. And at least 30 or so updates per second if you want to go smoothly and not have people bitch about jittering. That makes Doom3 look like it runs cheap in comparison. It's flat-out wasteful, your CPU time is better spent on real-time calculations.

Even if dynamic lighting isn't always necessary from a gameplay or visual standpoint, it is immensely useful in streamlining development environments. Companies used to sometimes have to run dedicated compile machines for hours (and, in rare cases, days) at a time. In the time you're spending compiling, I could rough out an entire new map or re-light an existing map 2-3 times over. And, as hardware advances, any perceived gains in lightmapped stuff will disappear entirely in favor of the smoothness of per-pixel details.

Seen the shots of HL2 textures in D3? They look better in d3 lighting than HL2 lighting. And that wasn't even a very good map. :razz:

---

www.button-masher.net

I thought when you had a changing light source prerendered (ie. the lightstyles in HL1), the compiler just calculated how much light from that source reaches each luxel, and the game engine calculates what colour that luxel should be realtime. Correct me if I'm wrong...

Oh, and anyway, why are you playing the same map for 24 hours? :razz:

---

"If you talk at all during this lesson, you have detention. Do you understand?"

• My yr11 Economics teacher

Dynamic lighting is the way of the future, I think. High system requirements or not.

It just makes more sense when comparing it to a real-life world, and that's what it's all about, isn't it ?

---

:: Forceflow.be :: Nuclear Dawn developer

all this talk about dynamic lighting is starting to convince me. ive changed my mind: real time lighting is the way of the future but for now i think it would be better to have a mix.

---

Bullet Control: $5000 for a bullet.
"I would blow your f**king head off! ...if I could afford it. I'm gonna get another job, start saving some money... then you a dead man!"

When you look at the top of the line in each area - Half-Life 2 for prerendered, and Doom 3 for dynamic, Half-Life 2 is the clear winner. The lighting in Half-Life 2 just looks better. Doom 3 its dark light dark light, and the real-time light is emphasized way too much. Half-Life 2's lighting is subtle and adds to the scene instead of taking over the scene. Perhaps that is a result of art direction and not technology, but I think the point remains - current real-time tech just looks "wrong"; the shadows are too dark, and so the games that use it compensate for the inability to produce realistic subtle effects by going with a high-contrast super shiny style.

---

http://maze5.net

Honestly, both together is a good option for right now. But we can't take current games and make a decision, since current machines aren't powerful enough. Which is why current real time lighting has such hard edges. Your computer can't continuously calculate the reflections and refractions that almost go on forever in real lighting. If they could: your opinions would change.

I think the addition of a physics processor would greatly help, since it can be used for anything you want (Yes, it can. You can even use your GPU to process sound if you write code for it) So, perhaps some companies would utilize it for lighting others for physics, or a combination. Who knows? Maybe there will be a duel core processor on graphics cards later so one can deal with light while the other draws.

The first and foremost problem as to why this stuff still isn't mainstreamed in games is the computer architectures that are out and affordable. Need to get rid of this bottle necked design first, then other things will be easier to do.

---

Blame it on Microsoft, God does.

Perhaps that is a result of art direction and not technology

Yep. Shadows are only as dark as you let them go. If you have pure black shadows, it's not the technology, it's you not filling it in (for whatever reason). That applies to most lightmapped engines as well (particularly when you get to stuff developed without CSG editors) - radiosity is not the standard.

---

www.button-masher.net

current real-time tech just looks "wrong"

You can't honsetly tell me that it looks more wrong than HL2's
shadows. Let me give you an overview of HL2's shadowing on models:

-Shadows only come from the sun, so pretty much all indoor lighting is completely borked

-You get shadows even when you are standing in the shade of a building
or something, shadows overlap each other, like when you hold 2 doors
over top of each other, the spot where they overlap will be darker than
the spots where they dont

-Models dont recieve shadows, especially not from themselves

As for a physics processor, I think it'd be more lucrative just to
perform those simulations on the gpu, considering that top of the line
gpu's now are just barely pushing their limits. Using the gpu for
doing some other stuff seems more logical than having to have another
card residing in my computer. The only problem is that it becomes
icky to formulate your task into gpu-terms.

---

To get something to work, sometimes you just have to beat your head against the wall longer; the skin grows back, but the brick doesn't.

Source hates soup!