Lovely Salsa
Banned
How was original super mario bros made? I always wondered this
Ancient photoshop?
Ancient photoshop?
-
Hey Guest. Check out your NeoGAF Wrapped 2025 results here!
How where when Metal slug sprites made?
- Thread starter dragonzdogma
- Start date
dragonzdogma
Member
Subpar Scrub
Banned
Don't be a smarmy jerk because your thread sucks.
tearsofash
Member
Don't be sorry. I honestly couldn't tell you where this belongs. It's video game art.
dragonzdogma
Member
Do you think they salvaged the 3d models tho?
The Real Abed
Perma-Junior
Bit by bit. Literally. There were only like 64 colors. They just used hex numbers for the colors they wanted and put them in a grid.
Lord Error
Insane For Sony
I've never seen this before and it's really interesting. What really intrigued me the most is that the two render planes are used for the main character and the enemies, and they constanly flicker between them. It's like the even frames of animation (or sync refresh) are drawn on layer one, and odd frames are on the layer two.
My best guess as to why this is done is that it makes it possible to render things twice as fast, as the process of drawing on these layers is something probably hardware accelerated, so while the graphics hardware draws on one layer, you prepare stuff on the CPU that's going to go to the other layer on the next frame etc.
I can't imagine that software developers would not write a simple tool for drawing sprites before actually coding them directly through hex numbers.
Lord Error
Insane For Sony
I know on C64 at least there were some sprite editors. You'd get a grid and you put colors in there, then it saves the sprite into the data stream that you can then use in your game. But early on, I'm pretty sure sprites were made using pen and paper grid, then entering numbers accordingly. I did that for a few sprites on C64, it wasn't that time consuming tbh. because sprites were small.
As for the backgrounds in those games, they were practically always done using reusable blocks that could be tiled, and sometimes color-swapped. An example is the infamous cloud = bush in SMB.
The Real Abed
Perma-Junior
They could have. It would have been proprietary and probably output it as code they could put in the game code as assembly.
I'm sure by the time SMB3 came along they had proper tools though.
they did. I'm not sure the exact editor they used for the early NES games, but this is the one they used for SMB3:
And this is an unreleased retail version of the SMS Graphics Board, which is similar to a sprite editing software and hardware they used for the master system:
Photoshop was definitely available, but SNK (and most likely Irem/Nazca by extension) used their own in-house software to create art assets. There probably wasn't anything magical about these tools though, mainly that they were tailored towards stuff like palette and sprite restrictions for the hardware they were developing for (Neo Geo in this case).
Based on the concept art that is out there it's not conclusive wether they scanned and drew over it, or if they just started created from scratch. Either way, there's an inhuman amount of talent at display, from everything to backgrounds to vehicle design to character animations. There's no "cheating" the way this game looks, it's pure craftmanship by people at the peak of their abilities.
CrispyBoar
Banned
You'll have to ask SNK themselves. 
Lord Error
Insane For Sony
Man, as it sways back and forth, the detail never loses its sharpness. Every line is perfect on every frame, there's not a sign of some unnatural looking pixel jitter anywhere, even though there's like a thousand animated components in just that one sprite...
Now imagine if you take a static image from that animation, load it in Photoshop and rotate by one degree. Lol.
dragonzdogma
Member
Man, as it sways back and forth, the detail never loses its sharpness. Every line is perfect on every frame, there's not a sign of some unnatural looking pixel jitter anywhere, even though there's like a thousand animated components in just that one sprite...
Now imagine if you take a static image from that animation, load it in Photoshop and rotate by one degree. Lol.
I might be wrong about this particular sprite, but I believe the neo geo could free transform sprites via code without any need for the artist to animate rotating sprites or angled sprites.
for example there is this one part on metal slug where you can use the tank on the very edge of a cliff and you will notice that the sprites are completely different when the tank is tilted.
Also some interesting sprites effects can be seen whenever the player chooses to continue, the character sorta scales from the sky to the ground into position.
If you see the bridge gif from metal slug snow stage, is something out of this world.
The neo geo has no support for sprite rotation.
dragonzdogma
Member
Okay, now I'm not sure if your familiar with nearest neighbor options on phothoshop
but if you rotate any metal slug sprite by one degree it would destroy the look, how is it possible for the neo geo to keep the sprite in it's actual uniformed sequence? 0_0
er, because they're drawing it all by hand.
The sprite you're looking at isn't many individual sprites being rotated and tweened, it's all one sprite that has many frames of animation.
dragonzdogma
Member
Sorry, perhaps I didn't explain myself properly,
I wasn't referring to any sprite in particular, I was talking about how if you rotate (say) the tank of metal slug via phothoshop using Nearest neighbor (only tech I know that keeps the pure pixels format) it destroys the sprite. but when the neo geo does the rotating (via code,using only one sprite) the sprite remains perfectly drawn with very little to no destruction of the actual place of the pixels.
I hope that made more sense :/
the neo geo isn't rotating. The neo geo has no ability to do that. They actually drew all the frames of rotation by hand.
but, to your general question - there are many ways to rotate an image using math, and different rotation techniques produce different visual artifacts. Nearest Neighbor is an interpolating algorithm for rotation - each rotation produces new interpolated pixels. This makes it considered a noisy (as opposed to lossy) scaling method. It takes the original image and adds new information to it through interpolation.
You can apply a pure, non-interpolated rotational matrix to pixel positions and not produce any visual artifacts at all if you wish. It all depends on the rotation algorithm one is using, and which algorithm you are using depends on many circumstances, such as speed and accuracy.
I don't think you understand what Krejlooc was saying. The Neo Geo didn't support sprite rotation the way you describe, only scaling. The reason why it looks "perfectly drawn" is because they drew all those frames (or used a rotate function in their art tool and then cleaned it up pixel by pixel, same difference).
Probably the most notable game to do this is the modern King of Fighters games.
Lord Error
Insane For Sony
I know what you mean, but even if NeoGeo had some hardware based sprite rotation support (which I don't think it did, it had scaling though) or if they used a CPU for sprite rotation like some Amiga games did - that would usually create some stray pixels around the edge that I don't see on that sprite. Still, I'm not excluding the possibility that the cannon tower was made to sway using a technique like that. Still, seeing how many other things are animated on that tower (like changing shadows etc) I doubt it.
I think that's what nearest neighbour does. It doesn't interpolate pixels creating blurry looking pixel art. Actually, surprisingly enough, that metal slug sprite rotated by 2 degrees in Photoshop using nearest neighbour, looks very respectable. Rotating it by 1 pixel however introduces some stray pixels around some of the edges - but only very few. I don't think it's entirely impossible that the gun tower on that lobster sprite was swaying using some CPU-driven sprite rotation (some Amiga games were able to accomplish that, despite hardware having no support for it - MC68000 had enough grunt to do it), although I think it's unlikely.
*edit* Oh wow - Nearest Neighbor does in fact interpolate, I've just witnessed it, it just doesn't interpolate the color, so there's no blurryness added. When you do free transform in Photoshop, the realtime preview of that is a simple matrix transform, which ruins edge lines of the sprite in many places. But when you apply the change, final transform looks way, way better, with most of those ruined edges nicely cleaned up. Although that makes it very clear to me that no kind of realtime rotation available to NeoGeo would be able to produce results that are as clean as the actual sprite we see above is. It's an open and shut case IMO that the rotation was manually animated, even if the big enemy sprite was a multi-sprite composition.
Very obviously layering of some sort.
It's a single layer sprite, and the number of independently rotating segments would be impossible to perform in software on the m68k inside the Neo Geo.
Now, what is likely is the sprite was drawn in segments and combined using something like deluxe paint, but each segment would be hand tweaked. This is not tweening.
Emperor Bohe
Member
Stabby McSter
Member
They also did this with the BlazBlue series, IIRC
SallyIsNow
Banned
For whatever reason this post made me lose my shit for like 10 minutes. :lol
Lord Error
Insane For Sony
While that may be so, I think the end result is rendered as a sequence of single layer images. You can see the rendered layers on the video posted earlier in the thread. Maybe it was just animated in a more layered fashion to preserve the sanity of whoever had to do it. To make the job at least a bit more bearable. Keep in mind also that the metal slug tank itself would animate using stretch and squish to some degree, so that was clearly done with every new frame having to be entirely different.
Nearest neighbor itself actually describes the interpolation method. It interpolates the final pixel output using a single "nearest neighbor" pixel to a considered pixel. This produces a minimal amount of noise (you get 1 final pixel for every pixel considered) but you still interpolate your data.
What I was saying is that there are numerous ways to rotate, including ways that don't interpolate at all. The problem with doing that is your rotated image produces holes when doing this:
This type of artifact is called Speckling. And it might be acceptable in certain situations. Really, the result of your rotation depends entirely on method and the method depends on the needed speed and accuracy.
Lord Error
Insane For Sony
Yeah, edited my post with similar findings just before you posted that. You can see practically both things at work in Photoshop, before and after you apply the final transformation. That makes me wonder which technique was in use by the arcade hardware that had built-in rotation.
SmokedMeat
Gamer™
Loved the sprite work in In the Hunt as well. Maybe even more then Metal Slug.
Very obviously not so. The only thing on Huge Hermit that is layered are the effects such as the water and blast. There are no layers on the other enemy objects (for their idle animations anyway). You're welcome to go through Huge Hermit's VRAM data which I've conveniently uploaded here and here, but having used that myself to piece together those animations, I can tell you it's not.
An example of something that does use layers though is the Ohumein-Conga's attack animation, which has a layered arm.
Also the Huge Locust's grabbing attack, which requires certain parts to be drawn overtop.
EDIT: Also like a million other people have said, the NeoGeo can't do sprite rotation.
Eolz
Member
I know it's not really what you're asking for, but SNK talks a little bit about it on their KOF XII website to compare it to their new DOT ART method that they used on KOF XII and XIII.
Hardcoding. There weren't that many pixels, and the hardware was limited, so they hardcoded the sprites in using binary, from my understanding.
Sega's old super scalar games apparently used inverse rotation, they would consider an array of pixels on the screen that need to be filled (X by Y) and figure the final pixel output based on an array of original, non-rotated pixels. The reason you get speckle errors in the first place is because you operate under the assumption that every pixel will run through a rotation formula and arrive at one final output location, but accuracy errors will sometimes produce pixels at the wrong spot. The speckles you see are pixels that are missing, being drawn in the wrong spot because of weird fractional errors. If you go backwards instead, you ensure no speckle, every pixel that should be filled is, and the process is fast and works itself well to being handled in hardware. It uses no interpolation.
CrazedProfessional
Member
If I recall the Sega Saturn version of Metal Slug came with a digital artbook full of unfinished sprites and pixel art.
Those SNK folks were frame by frame, pixel by pixel for sure.
Those SNK folks were frame by frame, pixel by pixel for sure.
erotic grandpa
Member
The gag in this one is so subtle. The artists definitely had fun with the Metal Slug games.
MrCunningham
Member
The sprites were created on paper, and then refined pixel by pixel using graph paper
from there they would code each pixel in one by one on a grid (I would imagine, going that far back to the first Donkey Kong and Mario Bros. 1).
Yeah, Mortal Kombat had this issue on the SNES as well. The sprites were designed for the Genesis resolution and ended up really fat on the SNES. for MKII on the SNES, they squashed the sprites horizontally so they wouldn't look as wide when stretched to 4:3.
Sprites at native resolution:
(Genesis left, SNES right)
4:3:
SNES:
Genesis:
The SNES sprites were lower detail. I think they may have also did the same for the sprites in Earthworm Jim 2 for the SNES.
The Real Abed
Perma-Junior
The Metal Slug games are impressive in how well they're animated and the sheer amount of frames they put on the cartridge. Though it was a NeoGEO game so the cartridge was bigger than a cinder block, but still.
sir_kevith
Member
lol never knew he pulled out a penis. Does the female monkey do the same?
Well I'm going to go play Metal Slug now after seeing those gifs. I've beaten all of the games up to XX but Metal Slug X was my favorite. Probably because it was the first one I beat.
Gif source (with tons more gifs): http://smg.photobucket.com/user/exa...GIFs/Video Game GIFs/Metal Slug?sort=3&page=1
Gif source (with tons more gifs): http://smg.photobucket.com/user/exa...GIFs/Video Game GIFs/Metal Slug?sort=3&page=1
CharminUltra
Member
This is awesome. Glad I kept reading (first few posts kinda turned me off the thread.)The sprites were created on paper, and then refined pixel by pixel using graph paper
from there they would code each pixel in one by one on a grid (I would imagine, going that far back to the first Donkey Kong and Mario Bros. 1).
Eh, thats the way people visualize the data because it's convenient, and matches the way it'll be output on screen, but that's not usually the case. The NES uses planar graphics, not chunky graphics, to save space. Where chunky graphics could be literally mapped in a grid, planar graphics must be spread across 2 8-bit bitplanes. This means a 4-color pixel can be represented in 2 bytes rather than 3 bytes.
Planar graphics work by treating each bit in a byte as a separate component of a binary number that is computed across n number of planes. so like so:
Code:
bitplane 1: 1 0 0 1 0 0 1 0
bitplane 2: 0 0 1 1 0 0 1 1to see the resulting sprite data, you read top to bottom, to produce 8 pixels:
Code:
b1: 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0
b2: 0 | 0 | 1 | 1 | 0 | 0 | 1 | 1
Re: 1 | 0 | 2 | 3 | 0 | 0 | 3 | 2where each number of "Re" maps to 1 of 4 palette colors. This means the two bitplanes must be stored as sequential bytes, represented in hex:
byte 1: 10010010 = decimal 146 = $99
byte 2: 00110011 = decimal 51 = $33
thus, every 8 pixels horizontal is represented as two bytes, and are organized into tiles that are 16 bytes long. Our row of pixels above would be $9933. That would be the data they would be inserting into the binary. If we wanted to make a single tile of the row of pixels above, our sprite data would be $9933 $9933 $9933 $9933 $9933 $9933 $9933 $9933. That's one 8x8 tile.
stargateheaven
Member
but isn't it original made with layers, then exported as 1 single sprite? edit: duh thats what you said. nvm. lol