@@Ashona I have started to work part-time in order to make more videos so I will upload more regularly now! :) (in my description I have explained now why I have been unable to post a lot of videos)
0:09 to be fair, Asteroids looked a lot better back then than what's shown here. It used vector graphics instead of pixels, so this low-res capture doesn't do it justice!
It looked especially good on the horizontal twin table top set up that you could find in some cafes. CRT monitor probably smoothed out a lot and enriched the look.
@@hpt08 again, the CRT didn't have to smooth out anything. It didn't have pixels but instead, the electron beam drew the shapes directly on the phosphor layer. It was basically infinite resolution!
@@Noel-yh4xp I grew up playing an old hand-me-down vectrex. They use an oscilloscope as a display and doesn't use polygons or pixels. Instead it uses vector based graphics.
@Coder Husk The game shown as illustration (Virtua Fighter) is one of the earliest 'fully' 3D games, and both in its original arcade version and its PC ports it used an unusual custom 3D accelerator chip design that was based on quads. One of those quirky things that happen when a market is new and everyone is trying everything. It's just funny that of all the 3D accelerated games, DigiDigger picked one of the few that _don't_ use triangles 🙂
It's a great video, but there is some misinformation. 3D graphics cards wasn't the thing that enabled true 3D graphics. Quake was released before those and it was a true 3D game with texture maps. A lot of other games used polygons before that as well. some were fully 3D, but were flat shaded. Others were a mix of 2D backgrounds and 3D characters. The game Outcast was released a few years after the first 3D graphics cards and it didn't utilize them at all. However it was one of the most beautiful games of that time. It used Voxels for the terrain and textured polygons for the characters, buildings etc. It also had things like anti-aliasing, depth of field, bump maps, character shadows. It was game ahead of it's time. So no 3D graphics cards (or 3D accelerators if we want to be accurate) didn't enable true 3D. It just made it a lot faster. It's also a bit ironic that the game shown on screen is Virtua Fighter. That game had a PC release in 1995 and it didn't use 3D cards.
09:20 - Missed opportunity talk about baked vs real time rendered shadows. Two really important phases in the history of game rendering. Baked shadows where shadow maps applied to the textures beforehand, to avoid having to do this in real time. Would have loved to see more info on shaders, and some of the more advanced renderpasses we see in modern games as well (some of the techniques that have progressed graphics from the N64 age of graphics to the graphics we see today) The video is an ok sweep through a few rendering methods and concepts, but you are missing some important milestones and you deep dive into things a bit randomly. It would have been better if you made a firm commitment to either go deep on the technical structure in each of the parts, or made it more of a history sweep with links to other deep dives. Also as you see from the comments have quite a few details and terminology wrong. Please don't take my criticism the wrong way. I liked the video, but felt there was room for improvement, especially since the idea of doing such a video (a history of rendering techniques) is great.
Binary Space Partitioning is not even close to what you've shown in the video. It's about compressing surfaces into a binary tree depending on their intersections, thus saving only equation parameters of the surface equation instead of actual vertex positions. What you show is more similar to an octree partitioning, but in 2d.
@@MrDavibu it's fine, since I remember it not that correct as well. To be more precise it's the way planes are sorted depending on what side they are in relation to the normal of the chosen plane. That's why it's called "space partitioning".
No. Confusingly enough "8-bit art style" refers to the "8-bit systems" art style, mainly the NES, that had 16 colors, and could only use 4 of them in each 16x16 tile. Meanwhile the "16-bit art style" is *actually* 8-bit because the "16-bit systems" like the SNES could usually display 256 different colors (or close to) on the screen.
Also. The “cel” from “cel shading” is not named because the quantized shades form “cells”, but is named after the cartoon shading it was meant to emulate. The shading painted, by hand, onto cellophane sheets or “cels” used in cartoon production.
There is a few errors i the video. Most notable... 8 bit era didnt have 8 bit grapics. It had only 2 bit grapics partitioned into sprites out of a pallet of usualy 16, some Times 32 colors. Even the 16 bit era mostly didnt have 8 bit grapics, but mostly 4 bit grapics put of a pallet of typically 16. Its also worth saying that most older consoles and also some computers in that era did have a 2d grapics chip that was driving the grapics. There was in most cases no cpu involvment in showing the grapics.. It was really mostly PC that lacked that ability... sort of. Pc the cpu renderad pixel by pixel. This pushed pc cpu to have more preformance partly to offset this disadvantage. This was why 3d exploded on pc.
8:35 Cel shading comes from traditional animation. From Wikipedia: "The name comes from cels (short for celluloid), clear sheets of acetate which were painted on for use in traditional 2D animation."
Yes, this is one of many strange assumptions and false statements in this video. It's a "ok" video, and a good general summary but the research is flaky, especially in terminology and several technical details, that are mentioned in various comments here.
@@elecblush Cel shading is just a slightly different term in 3D graphics... It's not a false statement or anything. You're making way to big a deal about this. Besides you could use the same logic to justify the name for the original style of animation techniques on hand drawn 2D animation.
@@HerbaMachina no. Its named so because of the shading technique's resemblance to animation cels. It's not an industry specific use of the term to refer to units in a color gradient. The original pixel art of 2d sprites isn't "cel shading" because while colored and animated like original 2d animation, it doesn't _resemble_ it, which is what cel shading is named for. Resembling old school animation cels.
I like this video but watching it as someone who's made 2D game engines is painful, most older games used palletted graphics, where a sprite would have 1 byte per pixel which would be an index on a table of colors(usually 3 byte), sprites would also have a palette index which would point to which palette to use, some early polygonal engines like Quake also used this but it stopped being used that much in the mid-late 90's
This is how I programmed all my own crappy games. The best of this technique was the ability to set the whole palette by calling an interrupt. You did not have to change the pixels to get animation. If used in clever ways you could change the palette and make for fast changes. I used this mostly to get lightning effects or even to control 7 segment displays on screen. I loved that era. It was easy to work with.
Some feedback: there is a static background noise. You sound like being in a huge room with echo (google fixes, but recording in a closet is a funny and partially crazy solution I heard). A lot of snappy sounds that can be fixed with a shield thingy (no clue what it is in english). Interesting video still! But sadly not something I like having on in the background.
Finally you are back. Yes, Graphics are very important. Visuals are the top 3 most important thing in the games. You should see Game-jam submissions. The graphics they make is something else in so less time.
Cels are the transparent sheets used in traditional 2D animation. The individual pieces of 2D art are painted on cels and then multiple cels are layered on top of each other to form a single image. Cel shaders are named this way, because they emulate the look of hand drawn toons.
LADS! HES BACK! The legend has returned to reunite us for a few precious minutes, but will soon return to the void again for an indefinite amount of time.
One significant change that was omitted was when fixed function pipelines went away and we started witing our own shader code. A lot of cool graphical effects came about because of this change. I think it was around 2005 but that's just from memory.
Ray-casting and ray-tracing are not the same thing. The main thing separating them being that Wolfenstein uses ray-casting for the sole purpose of determining whether or not a pixel intersects the mesh (consider this to be 1 bounce as the first cast is the final result), whilst ray-tracing aims to determine the path of a light ray as it bounces off surfaces before reaching the camera (usually making use of several ray-casts to achieve because light rays tend to bounce off multiple different surfaces each with their own unique qualities). Other than that, it's a great summary considering you managed to recollect 40 something odd years of game render tech development in the span of 13 minutes!
I would have liked more detail on how these systems make sure nearer surfaces are drawn “in front of” more distant ones. Back in the day there was the idea of “depth buffers”. Did old doom and/or modern cards use these?
It really irks me that you have used filtered/upsampled (2x / 3x) outputs from the 8bit and 16bit games instead of raw pixels. I know most retro interested viewers know this is not the resolution games were rendered in original, but when your video is about historical context, it would be better if you used more historically accurate representation in your examples.
1:59 "8bit artstyle you know from early Nintendo games" Wrong. The NES was even more limited than that, using 2 bits per pixel. It had 56 hardcoded colors, which you had to choose from to form 4 palettes of 4 colors for backgrounds, and 4 palettes of 4 colors for sprites. You could also choose a color to be transparent.
I know this is just a very simplified overview and maybe this is just too pedantic of me, but is anyone else bothered by the fact the video goes from making a big deal about rendering moving toward the use of triangles only to cut to Virtua Fighter, a game that famously opted for quads as the basic primitive? (Both on Saturn, pictured here, right? And on the Model 1 arcade board)
Well I didn't really want to overcomplicate things introducing quads as well, but you're right. It still shows what early 3D graphics with flat shading looked like in action, even though it uses quads and not tris (:
@Coder Husk The GPU the Saturn uses is quite literally famous for rendering in quads at a time where a lot of other GPUs were abandoning that approach for tris. No amount of jerking yourself off about it will change that. The only way to get a tri out of a quad requires you to do calculations for a 4th redundant vertex, which is computationally wasteful and results in terrible performance. There is no native support for tris on it, or the later released Nvidia NV1.
Did you just totally skip over 2-bit and 4-bit colour? You've fallen for one of the classic blunders! What's usually considered 8-bit art is not art that uses 8-bit colour; it's art that was suitable for 8-bit computers. The NES for example has only a 64 (6-bit) colour palette, and at any one time it can only display 13 of those colours. That's just a fraction of the 256 colours you can display with 8 bits yet we call those 8-bit graphics. I'm also a bit disappointed that you didn't mention any of the math used for actually figuring out where objects get placed on screen or any details about the core rasterization algorithm that turns those mathematically perfect triangles into imperfect pixels... though I guess it's fair since showing a bunch of math probably isn't that good for audience retention :P
Cartoon rendering: Cel is NOT a "color cel". "Cel" comes from animation. Early animation was drawn on paper and inked and painted on celluloid sheets for photographing. These sheet were called "cels" for short. The colors used in cartoons were just "paint". Cels were transparent (five was the practical limit) so you could stack cels, allowing production to only animate and photograph the parts that moved. This info brought to you by a guy that made commercial 3D cartoon rendering engines in the 1990s.
Imagine the reality we live in needs to refresh and reload like video games 😂😂😂 Grandparents: You're fortunate to be alive in this time my grandchild. When I was your age we had to wait much longer for the world to refresh or reload in order to say hi to our friends across the road.
Great video, but you kinda skipped how 3D rendering works in first place. How are the triangles stored? How is depth handled? How does the GPU convert different triangle postions to 2D pixels on my screen?
@Coder Husk thanks for the answer but that still doesn't explain what a 3d projection algorithm is, the math behind it, and what the GPU does. Veritasium goes very in-depth with how DOOM fakes 3d but never talks about how "real" 3d works, only what methods are used to further the illusion
Ohhh that makes a lot of sense. Looks like I need to sharpen my linear math skills to understand this better. Too bad this wasn't in the video, because that's really how games are rendered!
@@Patacorow Linear algebra is essential to understanding 3D rendering. Vectors, dot product, cross product, vector length, normalization. Matrix multiplication, transform matrices, matrix transpose and inverse. You don't really need to know how these things work, but what they represent in a three-dimensional space. For example, the dot product is a replacement for angles. You rarely use angles, nor the trigonometric functions sin/cos/tan.
Hmm. Ok, so you give a reasonable explanation of early ray casting but don't even mention what a vector is... Then you say "it's triangles" - does that really helpfully describe an obj file - which really isn't a complicated thing. And you've not even given any idea about Matrix multiplication in raterisation... You do kind of hint at the dot product so kudos for that. The issue is that early games aren't the simplest ways to do 3d graphics. They had to be super clever to do it with such limited resources so I don't think a chronological approach lends the simplest explanation. I know this was meant for a lay audience but... You never really explain the fundamentals. I don't want to be too harsh, it was a nice video but I think it was a missed opportunity. Tell you what, I'm going to make a video myself. And get 20 views. Hehe.
Good intentions, but this video was a big mess. No word about Quake? BTW, there were 3d games, before 3d accelerators were a thing. Also, a major leap was bi-linear filtering, missing that too. Next time, limit your subject and investigate that closer.
Note that the explanation of the name "cel shading" (around 8:40) is incorrect. The name comes from "animation cels", the sheets of clear cellophane used for hand-drawn animation.
This video ends waayy to soon. You did great explanation in what order things were drawn in doom or quake, but int in 3d world (and diferent types of them like Forward vs deferred) Also how Different texture types changed in time from simple photo like image to various PBR types like albedo/roughnes/metalic.
You didn't mention it, but it is perfectly possible to render true polygon-based 3D graphics in real-time without any GPU hardware support - just do all the computation on the main processor. Many early polygonal 3D games offered a software rendering option, including such titles as Thief: The Dark Project, or MiG-29 Fulcrum. The reason Doom devs choose raycasting was probably due to performance concerns.
10:47 UV maps in fact have nothing to do with reflections in rasterization. The only parameters needed to sample a cubemap as a reflection is the eye vector (pointing from the pixel to the camera) and the normal vector (pointing out from the surface). Using those, you can calculate the reflection vector by which the eye vector will reflect off the surface with said normal. This reflection vector is then used to sample the cubemap. I also want to note something about 8-bit palettes. It's not true that 8-bit computers and consoles necessarily output 8-bit color. A lot of them were much more limited. The NES for example has 64 colors available, of which only around 55 are unique (exact number is up for discussion). Additionally, the NES has 4 palettes for backgrounds and 4 palettes for sprites limited to 4 colors each, of which one color is shared between the background layer and one color is always transparent on the sprite layer, making for a total maximum of 25 colors on-screen at a time. Each palette is applied to a 16x16 pixel area on background layer and per sprite on the sprite layer (each sprite being either 8x8 or 8x16 pixels in size depending on the game). Another thing I'd like to add as a contributing factor to why graphics are as realistic as they are today is the fact that render pipelines as well as certain stages of rendering are completely programmable on modern graphics cards. Early 3D rendering hardware had more static feature sets and limited customization. Programmability gives graphics programmers a lot of power to customize how rendering is preformed by writing programs that run on the GPU known as shaders. This programmability is the sole reason why screen-space effects such as SSR and SSAO even exist, as well as the plethora of anti-aliasing techniques that most of them rely on image-processing.
It's somewhat misleading to connect 256-bit graphics to NES, as it used 2+2 bits for color, allowing for total of 16 different colors on the screen at the same time (actually even less as 4 of those were black/transparent duplicates). Some creative coding could be used to have more colors - that's the beauty of developing for these old systems with high restrictions. Other than that, a great video!
What dude..u are saying the applications of what we studied in college...I never knew binary tree will be an application of player tracking in a map...and I didn't even thought it would be like that...
The part about 8-bit graphics was WAY off... You did describe one way (out of many) in which to represent colors using 8 bits, but that's absolutely NOT what machines like the NES, Master System or Commodore 64 used, specially considering that these systems have vastly different graphical capabilities, despite all of them being 8-bit. Being 8-bit simply means that they have an 8-bit CPU, it has absolutely nothing to do with graphics. In conclusion, the expression "8-bit graphics" doesn't mean anything in a technical sense, it just references the general graphical style used back in the day of 8-bit machines.
Does the ability to discern a "3D" polygonal render from a "2D" render require a light-source? I know it's possible to view 2D renders successfully without a light-source being considered, but is a light-source required for viewing 3D successfully? Are 3D polygons without a light-source and without any shaders just 2D effectively?
Like the look of games with a very bold style however techniques like HDR, and Ray tracing add to the look of a unique style, I think artistry triumphs over raw horse power. This can be proven with Paint, I can take a canvas of 1024p and draw random shapes every where and it would objectively have a higher graphical fidelity of games like Hyper light Drifter or Dead cells it doesn't make it look better. Wolfenstein uses Ray CASTING, not ray tracing.
For God so loved the world, that He gave His only begotten Son, that whoever believes in Him shall not perish, but have eternal life.(John 3:16) if you confess with your mouth Jesus as Lord, and believe in your heart that God raised Him from the dead, you will be saved; (Romans 10:9)
This channel is a gem , suggestion for next video: scriptable rendering pipelines ( deferred, Forward, clustered etc) also how micropolygon rendering works ( I know geometry renderer have complexity ) but what about these new "geometry primitive shaders" as of this next gen rendering library have to offer and why developer don't use screen space micropolygon mapping ? before I know each engine engineer use different methods i.e GI systems have so many different capabilities. but they trying to achieve the same goal realistic rendering . Now theses new techniques use same principles of offline renderers ( i.e in movies & pre rendered cinematics) but multiple implementation to optimize the performance, however.. I think after ray tracing technology (and similar micropolygon mapping) new trend like adaptive path tracing and voxels rendering techniques use to archive much better quality ( like texture quality in Dreams& little big planet) you mentioned Dom use ray marching which is the same terminology for rendering equation of voxel engines .( Aka volumetric rendering) but I don't understand how shading in voxels works , it should not trace the final color as they don't have direct information of the texture projection unless from the G buffer ( deferred) . or "fake Phong shading" correct me if I'm wrong , Forward rendering is about geometry rendering aka vertex points , while deferred is texture projecting in the G buffer. ( Hence doesn't represent true 3D word) instead hybrid approach used like the scriptable rendering pipelines in unity and most recently Amazon game engine. there's some other techniques like forward+ and clustered ( voxelized GI) etc.. but these aren't new . just hybrid rendering for certain tasks & stages of the image output. I would like to dig deeper into how engine handle the scriptable pipelines .
3:03 Knowing what I do about the fourth dimensional construct housed in human vessel, referred to as simply "John Carmack," it's absolutely within the real of possibility
The use of *"ray tracing"* as a term to describe Wolfenstein is a bit misleading because the "tracing" part refers to reflecting the trajectory of light traveling through space, which is what those modern video cards do. What they did here is called *ray casting*, which is just checking for intersections, in this case with tiles in the binary top-down level maps describing whether a space is "off" (empty, hollow), or "on" (solid).
At the past with 8-bit graphics: "Woah the Graphics looks so realistic" Present with 4k graphics: "BoHohOHo the GrAphIcs LoOkS sO uGlY! is ThAt a PiXeL?!"
