Best software to create Doom/Duke3d type sprites?

As far as I know all monster sprites on Doom were real clay figures, but other games like Duke3d or Doom64 had high res 3d models, then pixelized and made intro sprites.

So of course im not going to create clay figures so im going to need to create 3d models, then pixelize them.

My question is: Which software do I use to create the models so they don't look "too complex" and look like 1995, and also, what is the best way to make them look like 1995 sprites? I like the Duke3d resolution better than Doom, I think Doom sprites look a bit dated but Duke3d still have a nice old school feel but with a bit increased resolution. So what would I need?

SLADE does actually have a basic pixel editor, but right now it's busted in the public betas. (The next one will fix it, or if you can build from source, you can have it now.) If you want to draw by hand, that's probably the easiest way, followed by stuff like GIMP.

I am not sure on Duke, but I know the id guys used a combination of Deluxe Paint II, and scanning actual models for the Doom enemies.

As for modeling, that's easy: Blender. The idea basically is that you'd create your model, take high-res grabs with that, downscale those grabs into sprites, and of course, put them into the Doom palette (unless you're not caring about making your models vanilla-compatible and/or color translateable, in which case you can keep them generally as either true-color PNGs or give them their own individual 256 color palette if that blocky look is desired). 

9 hours ago, Dark Pulse said:

SLADE does actually have a basic pixel editor, but right now it's busted in the public betas. (The next one will fix it, or if you can build from source, you can have it now.) If you want to draw by hand, that's probably the easiest way, followed by stuff like GIMP.

 

I am not sure on Duke, but I know the id guys used a combination of Deluxe Paint II, and scanning actual models for the Doom enemies.

 

As for modeling, that's easy: Blender. The idea basically is that you'd create your model, take high-res grabs with that, downscale those grabs into sprites, and of course, put them into the Doom palette (unless you're not caring about making your models vanilla-compatible and/or color translateable, in which case you can keep them generally as either true-color PNGs or give them their own individual 256 color palette if that blocky look is desired). 

But what is the best way to downsize? for instance in paint.net, you can enable smoothing or disable it. And im sure depending on the software it has different algorithms when downsizing. Also what should the size of the source screenshots be?

1 hour ago, mankubus said:

But what is the best way to downsize? for instance in paint.net, you can enable smoothing or disable it.

Ideally you downsize by percentage scaling.

The key idea is to make sure you've got plenty of crop space around your model in all areas (above, below, to the side), and to take your shots from a fixed view. Ideally all your shots for all your models are done from this same fixed view, for consistency's sake.

Like, let's say you take your source model shots at 2560x1440. Reducing that down to 10% would give you 256x144, with no blur from a fractional downscale - 10 pixels in the source image becomes one pixel in the reduced one.

Mind you, that won't (usually) be the actual rez of the sprite, just of the grab. You'd then begin to crop and remove unnecessary data for the actual, final sprite. The heights/widths of final sprites differs, but the Cyberdemon is (naturally) the tallest at ~107-112px tall (not counting the explosion frames when it goes up to the 130s), and the Spider Mastermind can be well over 230 wide in some of its frames.

The id guys used to put them into sheets like this.

From there it was just a matter of palette adjustment (the id guys did this via the famous Fuzzy Pumper Palette Shop), manual cleanup, and so on.

1 hour ago, mankubus said:

Also what should the size of the source screenshots be?

Doesn't really matter as long as you follow the rule above - try to have it be on a five-or-ten. In other words, 10% is okay, 15% is okay, 20% is okay, 12% is bad.

Generally speaking though, the cleaner the source image, the less manual cleanup you will have to do once downscaled.

I am not sure why one would render sprites at a higher res only to downsample them later. It'd be far easier, and probably yield better results, to render the sprites at the correct res in the first place. One thing to keep in mind: Event at 2x Dom res, humanoid sprites will become a pixelated mess, especially in the face area. Note how Doom's sprites actually have a lot of hand-painted pixels in them, like faces. Also note that the clay figures that were used for some of the characters have big extremities that can carry more detail at lower resolutions.

3 minutes ago, Cherno said:

I am not sure why one would render sprites at a higher res only to downsample them later. It'd be far easier, and probably yield better results, to render the sprites at the correct res in the first place. One thing to keep in mind: Event at 2x Dom res, humanoid sprites will become a pixelated mess, especially in the face area. Note how Doom's sprites actually have a lot of hand-painted pixels in them, like faces. Also note that the clay figures that were used for some of the characters have big extremities that can carry more detail at lower resolutions.

Because if you do them at-rez and it looks ugly for some reason, you can always fix up whatever caused the problems if you got higher-scale sprites. Without that, you're manually fixing every sprite because all sprites will have that problem.

Look at how much had to be done on the Revenant pictured above, for example. Now multiply that by rotation frames and so on.

Pixel Art Shader - Feedback Wanted -- this is what's behind Beyond Sunset by the way

Blender MD3 Exporter Rewrite -- for Blender 2.79

Doom Sprite Batch Render addon for Blender 2.9+

15 hours ago, Dark Pulse said:

Ideally you downsize by percentage scaling.

 

The key idea is to make sure you've got plenty of crop space around your model in all areas (above, below, to the side), and to take your shots from a fixed view. Ideally all your shots for all your models are done from this same fixed view, for consistency's sake.

 

Like, let's say you take your source model shots at 2560x1440. Reducing that down to 10% would give you 256x144, with no blur from a fractional downscale - 10 pixels in the source image becomes one pixel in the reduced one.

 

Mind you, that won't (usually) be the actual rez of the sprite, just of the grab. You'd then begin to crop and remove unnecessary data for the actual, final sprite. The heights/widths of final sprites differs, but the Cyberdemon is (naturally) the tallest at ~107-112px tall (not counting the explosion frames when it goes up to the 130s), and the Spider Mastermind can be well over 230 wide in some of its frames.

 

The id guys used to put them into sheets like this.

 

 

From there it was just a matter of palette adjustment (the id guys did this via the famous Fuzzy Pumper Palette Shop), manual cleanup, and so on.

 

Doesn't really matter as long as you follow the rule above - try to have it be on a five-or-ten. In other words, 10% is okay, 15% is okay, 20% is okay, 12% is bad.

 

Generally speaking though, the cleaner the source image, the less manual cleanup you will have to do once downscaled.

Impressive. So they went pixel by pixel recolouring the stuff? it must have taken an insane amount of time.

I would never be able to go pixel by pixel to recolour and polish the stuff. I would need some sort of recolouring brush or something because im not an expert artist to understand where each pixel should go to keep the correct shadows and stuff as you recolor which would take constantly changing colors to get the right different tonalities.

10 minutes ago, mankubus said:

Impressive. So they went pixel by pixel recolouring the stuff? it must have taken an insane amount of time.

 

I would never be able to go pixel by pixel to recolour and polish the stuff. I would need some sort of recolouring brush or something because im not an expert artist to understand where each pixel should go to keep the correct shadows and stuff as you recolor which would take constantly changing colors to get the right different tonalities.

That is precisely what they did, yes.

17 hours ago, Cherno said:

I am not sure why one would render sprites at a higher res only to downsample them later. It'd be far easier, and probably yield better results, to render the sprites at the correct res in the first place. One thing to keep in mind: Event at 2x Dom res, humanoid sprites will become a pixelated mess, especially in the face area. Note how Doom's sprites actually have a lot of hand-painted pixels in them, like faces. Also note that the clay figures that were used for some of the characters have big extremities that can carry more detail at lower resolutions.

Say you take two pictures of the same object, one at 20 megapixels and one at 10 megapixels, in such a way they appear to be identical -- same camera angle, same lighting, same lens, etc. If you downscale the 20 Mp picture so that it is the same size as the 10 Mp one, the downscaled picture will still have more details than the one taken at 10 Mp.

Why it works is because you are capturing more details at 20 Mp, and even though by downscaling you're creating a picture with the same resolution as if it were taken at 10 Mp, some of the extra details of having captured it at the higher resolution are preserved. This is incidentally how antialiasing works; a game is rendered at twice the monitor's resolution (or four times, or eight, sometimes more, I believe), and then downscaled. This results not only in less jagged edges, which is the most noticeable effect, but also in an image with more visible details than one rendered at the monitor's resolution.

18 hours ago, Dark Pulse said:

Doesn't really matter as long as you follow the rule above - try to have it be on a five-or-ten. In other words, 10% is okay, 15% is okay, 20% is okay, 12% is bad. 

That is not right. This will only work if the image's width and height are both multiples of ten. If you have an image that's 128x126 pixels, scaling by 10% will result in the value of 12.8px for the rescaled image, which will probably end up being 13px. The same image can be downscaled by 12.5% (which is the same as diving by 8) and that results in 16, which is fine.

If you don't want a fractional rescale, what you need to do is pick an integer both the width and the height of the original image can be divided into with no remainder, which will not always be possible. Either way I don't think that really matters all that much.

I was wondering how did they create Wolf3d sprites? never seen clay models for these.

Another helpful tool: 

This will create 8 or 16 angles of render images of a model that you can use for sprites, probably the closest you can get to Doom's style without making clay figures or doing some serious pixel pushing. The caveat however is you need to use Blender :)

EDIT: NVM, Gez beat me to it!

On 6/15/2021 at 6:18 PM, mankubus said:

I was wondering how did they create Wolf3d sprites? never seen clay models for these.

Entirely by hand, pixel by pixel.

23 hours ago, Dark Pulse said:

Entirely by hand, pixel by pixel.

That must have took a lot of time, I hope they got well paid.

2 hours ago, mankubus said:

That must have took a lot of time, I hope they got well paid.

Considering most of those same people made Wolf, Doom, and Quake, they got paid very, very well. :P

On 6/19/2021 at 1:46 AM, mankubus said:

That must have took a lot of time, I hope they got well paid.

I mean that's how most games back then were had their graphics done...

You didn't really have to go "pixel by pixel" per se though - Stuff like "Deluxe Paint" allready existed and offered a lot of tools that made the process of creating sprites "by hand" more easy.

On 6/12/2021 at 5:10 PM, mankubus said:

I would never be able to go pixel by pixel to recolour and polish the stuff. I would need some sort of recolouring brush or something because im not an expert artist to understand where each pixel should go to keep the correct shadows and stuff as you recolor which would take constantly changing colors to get the right different tonalities.

Advice from spriter: you can use adjustment tools from graphics editors such as the Curves tool and Exposure tool (both found in multiple graphics editors) to help you balance lighting, coloring and shadows to make sprites look more natural.

On 6/16/2021 at 8:18 AM, mankubus said:

I was wondering how did they create Wolf3d sprites? never seen clay models for these.

The Wolf3D sprites were all done by hand, which is precisely why clay and latex models were created for the Doom games - it was such a pain in the ass to do that they needed all the shortcuts they could get!

Duke3D's monster and player sprites were 3D models rendered in some primitive, NURBS-based modelling software that I forget the name of. Many textures were photo-sourced. Some of the weapon sprites were based on photos (the shotgun was a real shotgun bought second-hand from a gun fair and heavily drawn over), and a lot of the incidental props were simply drawn by hand in Deluxe Paint.