default header

Theory

Are game developers getting slower?

Moderator: JC Denton

Are game developers getting slower?

Unread postby icycalm » 12 May 2022 19:05

Are game developers getting slower?
https://www.patreon.com/posts/66343144

Image

icycalm wrote:Usually, I have all the answers, but this one stumps me, so I'll leave it here and you can tell me if you've any ideas.
User avatar
icycalm
Hyperborean
 
Joined: 28 Mar 2006 00:08
Location: Tenerife, Canary Islands

Re: Are game developers getting slower?

Unread postby SriK » 25 Jul 2022 04:20

I posted a long reply to the discussion in the Patreon comments. Hopefully it helps shed some light on this topic:

SriK wrote:I think there were a lot of good points raised in this discussion, but one very important point which I don't think anyone properly considered is the role of asset complexity. Take a look at the following two screenshots.

Heroes of Might and Magic:

Image

Songs of Conquest:

Image

There are simply far more unique art assets on the second screen, and more detailed assets at that. HoMM recolors its hill tiles into mountain tiles, recolors its roads into rivers, and I think even the sand on the side is recolored based off the grass; smart asset reuse, in other words, which however seems unacceptable in a modern game. So the Songs of Conquest guys have to actually draw the rivers and mountains, plus make like ten different varieties of trees (instead of recoloring one or two tree assets and trusting the player's imagination to fill in the gaps), and then they have to make countless different varieties of buildings and structures, PLUS animation on top of all this, PLUS develop their own custom lighting systems tailored for pixel art -- all so that the modern player can finally believe that they're managing a beautiful village in the middle of the untamed wilderness, instead of a bunch of simple "building assets" placed on top of a bunch of simple "ground assets", which is what the original HoMM would look like today to most modern players if it was re-released exactly as it was by an unknown developer under a new title. I'm not saying they're right to think that, only that it's what they would think; and the same applies for the original Civ too.

You might object: isn't pixel art ten times faster to make today than it was in the '90s? Shockingly, I don't think it is. Deluxe Paint seems to have been the cutting-edge pixel art tool of the '90s (I think even the Japanese used it) and the feature list looks almost the same to me compared to modern pixel art programs. There's only so much you can do to speed up a process like pixel art; ultimately the bottleneck is that it's pixel art, and at the end of the day some artist has to go in there and place almost every pixel one by one, frame by frame, character by character. Now of course the pipeline AROUND pixel art is much faster today, since you can drop an asset directly into Unity or Unreal once you're done with it and set an animation up in 15-30 minutes, instead of running around the office with floppy disks like they had to do in the '90s. But from what I know, I would say it amounts to a 2x or 3x speedup, rather than a full order of magnitude. Pixel art is still easier and cheaper than HD hand-drawn digital art, since the resolution is smaller, but that doesn't necessarily mean it's easy and cheap.

So, the art process is twice or thrice as fast; but the assets are say eight times more numerous; so why not hire four or five times as many artists as the '90s teams, to get the game out in roughly the same amount of time? That's how you bankrupt your studio and guarantee your investors a loss. I think this would be a suicidal move for most small studios, since they aren't operating with the level of financial freedom where they can shake off a giant loss, let alone several in a row -- exactly because they're working on minigames, as you put it, which are not always guaranteed to be major hits. So in the end many studios proceed understaffed, working within very tight budgets and team sizes.

Or for another example, take my game Steel Assault. Six years to make; now granted, only two of those years were full-time and truly dedicated to it, and those two years probably accounted for 75% of the development. (The rest of the dev cycle I was in college, then grad school, then working in finance to make enough money to go full-time, etc.) But even if you accept this, why the hell did a 30-minute arcade game take two years of full-time work?!? A game which could have been churned out four or five times faster in 1995 Japan?! And which ultimately looks far worse than, say, Metal Slug? The answer is that while the old devs could spend untold amounts of money on each cutting-edge game (Super Mario World's budget was in the millions, for example) modern devs are lucky to spend a fraction of that. Metal Slug took at least a full year to make, and that game had six or seven highly paid artists working on it full-time, probably more than full-time, moreover highly trained and insanely hard-working Japanese artists who had years and years of previous experience working as a team in person at prestigious companies like Irem. Meanwhile my studio had to make do with a cobbled-together remote team of extremely-part-time artists who had never worked together before, and whose art styles often even clashed with each other. Granted that my next game is in a better situation now, and that's one of the reasons why the art looks a bit better than the previous game, and that's also one of the reasons why it will not take six damn years this time around lol, but even now my studio has nowhere near the talent of the old ones, and may not for a while.

(And that's not to mention that you can hardly find ten currently active pixel artists today on the level of Nazca's best, let alone get them all on the same team working full-time for you on a project which has a chance of making back your investment. All the original Metal Slug artists are either retired or impossible to find, except Akio, and I asked him and he isn't interested in doing pixel art anymore.)

Just to be clear, all the above analysis only applies to games with respectable art, not atrocious looking shit like I dunno Super Cyborg which should ideally take three months to churn out. And even that doesn't happen; those games ALSO take years because the devs don't commit to them full time, and they're even smart for not doing so, since there's no way the game could make enough money to justify a career move like that.

Other than asset complexity, another factor is that games today are released on many platforms simultaneously, not just one. I think Recap sarcastically calls this the "age of ubiquity"; sarcastically because rather than finishing a game and immediately moving on to the next project, devs have to spend several months AFTER they've almost finished the game to port it, and then coordinate with each platform's managers and QA teams to fix platform-specific bugs, add platform-specific features, and plan a simultaneous release in order to guarantee a profit. This doesn't apply to Songs of Conquest, which looks to have been released on just PC and Mac, but it's true of many other games. With Steel Assault, for example, it took about three months after the game was effectively done to port it to Switch and release it; and in those three months, the Kazuma Kujo of 1995 would have come up with a whole new design and seen it halfway to completion, since his cutting-edge studio had the funds to work on many games at a time. (SA would have been a financial failure without the Switch port though, so the port was still worth it in the end, and there will be more ports and platform deals coming!)

As for the programming and design factors... I don't know, it seems fishy to me if these are the major bottlenecks. If you wanted to program a console platformer in the '90s, you had to be some sort of math whiz just to get the screen scrolling. And if you wanted to play music WHILE the screen was scrolling, you had to buy textbooks on assembly language and rifle through tomes of obscure hardware documentation. The same with every other genre and platform, more or less, correct? If programming is the bottleneck on the Songs of Conquest team, then I'd bet it's art- and effects-related programming, not the programming of the mechanics or the basic rendering systems, which should be easier than it was in the '90s. On the other hand, networking and online multiplayer are a major component of modern games, and THAT is still a giant headache which takes ages to get right, if the teams are determined to get it right. And on the other hand, programmers usually demand a higher salary than many artists and musicians (I think it should be the reverse, but it's what it is). So maybe these understaffed teams don't have enough of them despite all the aforementioned. And as for design, most of the major design problems in this genre have been solved already for the SoC devs, right? It's not like they're building HoMM from scratch; they already know what HoMM is like and how it plays, and they can copy it. The guys who made the original game didn't have that luxury.

So, in conclusion: while it could be true that Magnus Alm is less talented and less hard-working than Sid Meier, and while it could be true that he's much less capable as a team director, there's probably more going on here in my view. Should SoC have taken six years to make? I think that six years is still ultimately an insane figure, and the game probably COULD have been made in half that time, but when I sit down and think about it, the long dev time seems more plausible to me. I won't sit here and pretend that Magnus Alm and I couldn't work harder than we do, but we have our excuses, and as you can see we have many of them :) One of which is writing long posts like this instead of working, but I wanted to write about this topic eventually as an editorial for my studio's website anyway.

P.S. One potential solution to the art issue, as I see it, is the development of advanced AI which can seamlessly generate concept art based on user prompts and "convert" between different art styles on request. That would super-charge development of videogames in ALL art styles, not just pixel art. This kind of thing already exists, in an embryonic but still impressive form (look up DALL-E 2 from OpenAI), but even the most cohesive art it generates is often kind of generic and soulless. Probably because the only way researchers can successfully train these systems is to throw as much image data from the Internet at the AI as possible, and thus the AI system can't develop any real taste of its own. But as researchers develop even larger AI models designed to gobble up even more art, this problem may be circumvented through pure brute force. At that point it'll be up to the user's tastes and talent to guide the AI system ("generate me a painting of my D&D character, in the style of Frank Frazetta, in a setting like this illustration, with a composition like this painting, and the lighting configured like this other painting, and then animate him like this GIF...")
User avatar
SriK
 
Joined: 05 Nov 2011 15:12


Return to Theory

cron