Using "super sprites"

[Exitus]

I understand - and accept - the argument that ROTT would lose its charm if 3d enemy models were used. So I had an idea? What about a real-time enhancement that immediately converts 3d models or even ragdolls to 2d sprite images as you play the game, with the same framerate? Same 2d animated images from 8 different angles, but the variations would be infinite, while still retaining the charm of the 2d sprite.

[Parkar]

Originally Posted by Exitus

I understand - and accept - the argument that ROTT would lose its charm if 3d enemy models were used. So I had an idea? What about a real-time enhancement that immediately converts 3d models or even ragdolls to 2d sprite images as you play the game, with the same framerate? Same 2d animated images from 8 different angles, but the variations would be infinite, while still retaining the charm of the 2d sprite.

That's actually a pretty neat idea. I am not sure if it would "save" the charm of ROTT though since to some degree it would still have the limitations of 3d models which is part of the problem with them to begin with. Would defiantly be cool to try out for a standalone game though.

The problem, IMO, is that the enemies and other objects are(most of the time) realistic.
The enemies are real photos of people, and not cartoonish like in Duke Nukem 3d.
Point is that if you want them to look good you need a damn good 3d modeler and skinner.
Which brings you to the 3d model to sprite process. It would be a waste to turn such a model into sprites, too much work has to be put in to make such a mode.
Further more, turning every angle and motion into a sprite would take a whole lot of space.
The standard soldier has 91 sprites, which translates into 23mb in high quality sprites.
A 3D model would have around 1mb for the skin(that's really high quality) and 200kb for the model and animations. That's more than 19 times smaller than a good model.
And an other huge problem is that each sprites needs to be loaded into the memory, which takes up a lot of loading time. With a 3D model this would be a lot faster.

So overall, a really good quality 3D model is IMO recommended, it saves loading times and space.

[Exitus]

Originally Posted by The Stinger

The problem, IMO, is that the enemies and other objects are(most of the time) realistic.
The enemies are real photos of people, and not cartoonish like in Duke Nukem 3d.
Point is that if you want them to look good you need a damn good 3d modeler and skinner.
Which brings you to the 3d model to sprite process. It would be a waste to turn such a model into sprites, too much work has to be put in to make such a mode.
Further more, turning every angle and motion into a sprite would take a whole lot of space.
The standard soldier has 91 sprites, which translates into 23mb in high quality sprites.
A 3D model would have around 1mb for the skin(that's really high quality) and 200kb for the model and animations. That's more than 19 times smaller than a good model.
And an other huge problem is that each sprites needs to be loaded into the memory, which takes up a lot of loading time. With a 3D model this would be a lot faster.

So overall, a really good quality 3D model is IMO recommended, it saves loading times and space.

I suspected this, which leads me to another crazy idea I had.

This will probably never happen any time soon (if ever), but imagine a revamp of ROTT's sprite system where: a live human steps into a room with 32 cameras recording from 32 different angles, dressed like a generic ROTT enemy; he goes through all the motions (i.e. dying, firing, getting hit by a bullet, running, walking, crouching) and there could be many variants on each factor (especially dying); instead of being photographs, this is recorded as film; the film is then chopped down to 8 frames per second, and compressed using the same compression technology that is used for compressing traditional movie files.

You could then add algorithims for gibs, blood, stains, bullet hits, bullet holes in walls and sprites, for an infinite 2d variety. There could also be a filter where certain colors on the sprites can change to different hues, maybe even mix and match simple things like faces, hats, hair, to create a large pool of similar but distinct enemy characters.

It's basically the same motion capture method they used for ROTT, on steroids. Instead of 8 different angles, there's 32. Instead of 2-4 frames per second, it's 8-16. Instead of one animation death for each enemy, there are dozens, based on how they are shot and with what weapon.

With today's computers being having exponentially more memory and space that they did back in 1994, this could theoretically be possible.

[Altered Reality]

I understand - and accept - the argument that ROTT would lose its charm if 3d enemy models were used.

I absolutely disagree.

[Neuro]

Sprites should be left alone and no models should be made. Period. Problem solved.

[Altered Reality]

Originally Posted by Neuro

Sprites should be left alone and no models should be made. Period. Problem solved.

Yeah, you're completely and absolutely right. And no high-resolution textures either. And no ports either: the only true way to play it is with a 486 PC with MS-DOS, any other way means raping the game.
This is the ROTT HRP topic. Go and stop the project.

/sarcasm

[Neuro]

Originally Posted by Altered Reality

Yeah, you're completely and absolutely right. And no high-resolution textures either. And no ports either: the only true way to play it is with a 486 PC with MS-DOS, any other way means raping the game.
This is the ROTT HRP topic. Go and stop the project.

/sarcasm

Nah, look, buddy, it's good to improve the game but there should be a limit on how much you change it. For instance, I think that making high-resolution sprites would be a neat idea, but making 3D models wouldn't be.

If you keep things in the same style but make them better, you're money.

[Tea Monster]

Originally Posted by Neuro

For instance, I think that making high-resolution sprites would be a neat idea, but making 3D models wouldn't be.

So you want to play Quake IV with sprites????

One thing you could do, would be to make accurate models of the sprites - not just game models, but realistic, high-poly models. When you have them, render out sprites of them. Pose your model and then turn on the realistic lights and fire up the high-quality rendering engine. You could set up the lighting so that you get a faithful rendition of your subject. Then you pose the model in the same poses that the sprite came in and render out the various views that match the sprite.

This means that there isn't the fall-down that most 3DR source ports suffer from - that being that there don't seem to be any OpenGL coders in the community.

That way you can have really realistic looking critters and people - nicely rendered in very high-res - so that they actually DO look real - but you don't have all that 3D thing that you don't like.

[Neuro]

Originally Posted by Tea Monster

So you want to play Quake IV with sprites????

I'm not talking about all games; I'm obviously talking about making changes to old games only. Read with a little more thought.

[du8d]

I hate 3d models and I love sprites. High resolution sprites are preferred.
No 3d models is good enough even with the latest technology.

The size of sprites won't be a problem because it do not need calculation like 3d models. Just read from disk, draw it. That is all. But 3d models need calculation and so they needs to be cached.

[filipetolhuizen]

I would sugest widescreen modes with the proper FOV first to avoid the stretched look on new monitors.