Good news. Robo Recall, one of the best looking games available on a VR headset today, actually cost a lot less to produce than the rumored $10 million.

It’s certainly a compliment to the small team at Epic Games responsible for the game that the notion of a $10 million production budget was so widely spread and accepted. Robo Recall is great, but, it turns out, not $10-million-great.

The Origin

The suggestion of a $10 million budget started with an interview with Epic Games CEO Tim Sweeney at last year’s VRX conference where he addressed the fact that Oculus had offered to fund production of game so that it could be published for free on their platform, saying that “Oculus is funding Robo Recall which has a budget that’s close to the budget of the entire first Gears of War (2006) game.”

That statement, taken with another in a 2007 Wired interview with Epic Games Vice President Mark Rein, who said that the original Gears of War had a budget of “somewhere between nine and ten million dollars.” This was very reasonably presumed to indicate that Robo Recall’s budget fell near the $10 million mark. A nugget of info that’s been spread far and wide in the VR space.

SEE ALSO
'Robo Recall' Behind-the-scenes: Insights and Artwork from Epic Games

Lost in Translation

But it seems that something was lost in translation. Epic’s Technical Director of VR & AR, Nick Whiting, who worked closely on Robo Recall, hints that the game’s budget was actually much more conservative. .

“I’d have to check with Tim on that,” Whiting said in a recent interview with Road to VR when asked about Sweeney’s remarks. “He may have been referring to something other than the $10 million Gears budget that some people inferred from another interview.”

Colloquial use of the terms “budget” and “funding” is a likely root of the misunderstanding; what Oculus paid Epic to create Robo Recall is likely very different than how much the game cost to produce. After all, the game is completely free, but Epic almost certainly would want some upside on what the game would have earned if it had been launched with a price attached.

Whiting understandably didn’t want to get into specific figures, but laid out the scope of the team that created the game.

“There were around 15 full-time employees for about a year [working on Robo Recall], with the team starting to be assembled right after GDC 2016 and building over a few weeks to full size,” he told Road to VR. “In addition, we had six contractors for gameplay programming and QA, as well as some art and audio help through outsourcing. We also had a few people we borrowed for other teams at Epic for a few tasks like rigging and narrative for short bursts.”

robo-recall-tal-botI also reached out to Sweeney to ask for clarification; he referred me back to Whiting’s remarks regarding the team size and added, “I think that’s the best comparable measure of the resources required to develop a game of the scope of Robo Recall.”

Ready Your Envelopes

With that information, we can rough out the true scope of the game’s production with a little back-of-the-envelope math. I reached out to some industry vets to check the following estimates against their experiences in game development.

Game Industry Career Guide cites sources putting game developer salaries somewhere between $70,000 and $130,000 annually, depending upon experience. To give our estimate plenty of buffer, let’s assume Epic, as a leading games company, is attracting and keeping top talent; so let’s take the top salary figure of $130,000 and bump it up to $150,000 as our guess for the salary of those who worked full-time on Robo Recall.

For 15 full-time staff for a year starts our estimate at $2.25 million. Whiting also says there were six contractors for programming and QA. The goal of hiring contractors of course is to save money over full-time staff, so let’s estimate the equivalent of a lesser (but still cushy) $125,000 salary for each contractor, and assume they were there for the full year as well. That adds an additional $750,000 to our budget, putting things comfortably at $3 million.

Then there’s the outsourced art and audio help, and a few people who were borrowed internally for short bursts of work. These two pieces were likely the least costly from a staff standpoint, but to keep our estimate safe, let’s assume it cost just as much as the six contractors, and slap another $750,000 on our estimate. That brings us to $3.75 million.

Another cost worth considering is equipment needed during development. Let’s again offer plenty of padding and assume that each of the 15-person team needed a brand new high-end VR-capable computer for $2,000 a pop and an Oculus Rift + Touch for $800 each ($42,000), and round that up to an even $50,000 for good measure. Now we’re at $3.8 million.

SEE ALSO
Facebook Founder Says Company Invested $250M in VR Content, Will Invest $250M More

You might guess that marketing would be a major cost as well, and though we’ve seen a little marketing action for Robo Recall, it’s not been a major push by most measures. That and, if we look back at the original interview with Mark Rein which established the $10 million upper budget limit for Gears of War, the context of the discussion was Rein saying how inexpensive it was to “make Gears of War,” because of the company’s Unreal Engine tech. His point would have been not been as salient if he wrapped marketing dollars into that figure, and thus my gut is to also not include a marketing budget in the Robo Recall estimate. If you really want to count it, slap another $1 million onto our $3.8 million estimate and we’re still comfortably under $5 million.

Surely there’s a number of other smaller costs not accounted for in this estimate (I’m willing to bet at least one pizza was ordered during the development of Robo Recall), but we’ve got the major pieces, and by keeping our back-of-the-envelope reasoning at the high end of every estimate, we will have hopefully well covered smaller unmentioned costs.

– – — – –

So, somewhere around $3.8 million to develop Robo Recall. That’s great news, because it shows the sort of high production value that can be achieved in VR with a small team on a modest budget that’s significantly less than $10 million, and—extrapolating from what we see in Robo Recall—it gives us an idea of what a AAA VR game could look like in the future with a real $10 million production budget. So here’s to hoping that the next VR project from Epic Games does end up with that budget.

  • polysix

    if it had cost $10 million for *that* then I’d demand a recount and probably look at cutting people’s handed in food bills. ;)

  • trekkie

    There has to be a cheaper way to create content. Otherwise game devs with game ideas will never be able to bring their ideas to life and only a handful of big studios can do so… and so this poor dev goes and works for one of them. I am thinking procedural: characters, environments, materials, sounds etc should be able to be created automatically by evolution/genetics and tweaked. All lighting should be realtime and worlds physics driven by default. A creative dev should be able to get into a VR editor and rapidly create worlds and gameplay… yes we do have those components available but its all disjointed and disconnected.

    • 9of9

      Hey, as it happens these are all things I work with!

      These things are disjointed and disconnected – broadly speaking – because those things tend to compete for the same computation resources and tend to only work in specific cases.

      Physics-driven worlds are great, most of the time the physics you see is only there for fluff. If you want physics to affect gameplay and especially if you want physics in a multiplayer game, that drives up both computation overheads (to prevent things from glitching out and ruining the game) and network overheads (to make sure physics is synchronised over network) way way up.

      Real-time lighting is good to a point, but hits the GPU immensely. The better it is, the bigger the GPU hit and there are still no one-size-fits-all solutions for the really pretty stuff like global illumination, so that you can approach baked lighting quality in real time. The more performant solutions like Enlighten, for example, usually mean you have to be more careful with your authoring and places a greater burden on the artists to make assets that light well with that specific tech. Different technologies are more suitable for different types of games.

      Content creation is a different bag altogether. There is a big push for proceduralism and procedural content creation tools, but the issue is that you generally end up with having some kind of trade-off between quality and breadth, and either way simply creating procedural systems to create assets for your specific project takes a lot of development effort.

      As an example, you can have a technical artist spend two months making a generic furniture generator and have it spit out some slightly different, blocky, simple couches and tables and chairs and stuff and then use that to populate your world, but they’ll probably look more like something from 2006 and won’t be particularly interesting.

      That technical artist can spend two months and make like a really good chair generator that makes really pretty, AAA-quality chairs in the style of your project.

      Maybe if she spends six months on that chair generator, you’ll have something that you can use across multiple projects and styles, but then you’ll probably spend a month or so just defining the parameters so it creates the right kinds of chairs for you when you start a new project that needs chairs.

      Or like, you can have an artist sit down for a month and make six chairs for you, which is probably all you’ll really need without players ever noticing there’s only six chairs in the game.

      Using evolution and genetics for arbitrary assets would be awesome, but that tech is nowhere near where it would be usable for production yet.

      So it’s like, all of these things are doable, but you really have to pick your battles. No Man’s Sky probably pushes the envelope as much as anyone does in terms of procedural generation, but obviously the creatures it creates aren’t really assets you could use in Gears of War, and they look fairly samey after a while on top of that.

      On the other hand, Ghost Recon Wildlands uses substance for semi-procedural texturing on a lot of its assets and that saves some work, so you can have a hundred different types of camouflage with their own wear-and-tear patterns applied to your guns.

    • David Melton

      There is a cheaper way. Its called Unity. You don’t see nearly as many indie developers working with Unreal which seems better suited for larger teams like this.

      • trekkie

        Not a start to a UE4 vs Unity war (I lived throught the java/C++ and Direct3D/OpenGL wars)… Unity still needs content to be imported in and the content creation is what I am speaking of. BTW you can work very fast with both engines (have u check out Blueprint?) – PROVIDED – content is all ready imported in. Hence I was suggesting the next level of procedural content creation which might ease that pain.

        • Full Name

          Both are fine for quick development. Next.

    • No Spam

      I’d love to see AAA studios create a sub-licensing business for art artifacts. Imagine as an indie being able to license the models of Paris that Ubisoft created for Assassin’s Creed (and re-used for Eagle’s Flight).

      The original developer could recoup some of its investment, while indies would be able to leverage high quality assets for far less than the cost of recreating them from scratch.

      • Foreign Devil

        Yes I don’t know why this doesn’t happen. . .. Every new AAA title kind of re-invents the wheel by creating all new assets. . THere really needs to be low cost massive libraries of assets and textures.. studios could earn more profit by selling newly created assets and studios can save huge amounts of time and money by re-using, modifiying, existing assets. . but so few studios do. .

    • 9of9

      Bleh, my original reply to this seemed to have gotten marked as spam.

      Here it goes again:

      Hey, as it happens these are all things I work with!
      These things are disjointed and disconnected – broadly speaking – because those things tend to compete for the same computation resources and tend to only work in specific cases.

      Physics-driven worlds are great, most of the time the physics you see is only there for fluff. If you want physics to affect gameplay and especially if you want physics in a multiplayer game, that drives up both computation overheads (to prevent things from glitching out and ruining the game) and network overheads (to make sure physics is synchronised over network) way way up.

      Real-time lighting is good to a point, but hits the GPU immensely. The better it is, the bigger the GPU hit and there are still no one-size-fits-all solutions for the really pretty stuff like global illumination, so that you can approach baked lighting quality in real time. The more performant solutions like Enlighten, for example, usually mean you have to be more careful with your authoring and places a greater burden on the artists to make assets that light well with that specific tech. Different technologies are more suitable for different types of games.

      Content creation is a different bag altogether. There is a big push for proceduralism and procedural content creation tools, but the issue is that you generally end up with having some kind of trade-off between quality and breadth, and either way simply creating procedural systems to create assets for your specific project takes a lot of development effort.

      As an example, you can have a technical artist spend two months making a generic furniture generator and have it spit out some slightly different, blocky, simple couches and tables and chairs and stuff and then use that to populate your world, but they’ll probably look more like something from 2006 and won’t be particularly interesting.

      That technical artist can spend two months and make like a really good chair generator that makes really pretty, AAA-quality chairs in the style of your project.

      Maybe if she spends six months on that chair generator, you’ll have something that you can use across multiple projects and styles, but then you’ll probably spend a month or so just defining the parameters so it creates the right kinds of chairs for you when you start a new project that needs chairs.

      Or like, you can have an artist sit down for a month and make six chairs for you, which is probably all you’ll really need without players ever noticing there’s only six chairs in the game.

      Using evolution and genetics for arbitrary assets would be awesome, but that tech is nowhere near where it would be usable for production yet.

      So it’s like, all of these things are doable, but you really have to pick your battles. No Man’s Sky probably pushes the envelope as much as anyone does in terms of procedural generation, but obviously the creatures it creates aren’t really assets you could use in Gears of War, and they look fairly samey after a while on top of that.

      On the other hand, Ghost Recon Wildlands uses substance for semi-procedural texturing on a lot of its assets and that saves some work, so you can have a hundred different types of camouflage with their own wear-and-tear patterns applied to your guns.

    • Foreign Devil

      Well that’s kind of like saying there has to be a cheaper way to create feature films. There is. . but it will be an indie or student film. Same with games.

    • 9of9

      This is so annoying. I keep trying to reply to this and Disqus keeps marking up my comments as spam. I’m completely confused. Anyway, going to try and break this up into a few chunks, maybe that’s what’s setting off the content filters? Who knows.

      As it happens these are all things I work with!

      These technologies are disjointed and disconnected – broadly speaking – because those things tend to compete for the same computation resources and tend to only work in specific cases.

      Physics-driven worlds are great, most of the time the physics you see is only there for fluff. If you want physics to affect gameplay and especially if you want physics in a multiplayer game, that drives up both computation overheads (to prevent things from glitching out and ruining the game) and network overheads (to make sure physics is synchronised over network) way way up.

      • 9of9

        (1) Real-time lighting is good to a point, but hits the GPU immensely. The better it is, the bigger the GPU hit and there are still no one-size-fits-all solutions for the really pretty stuff like global illumination, so that you can approach baked lighting quality in real time. The more performant solutions like Enlighten, for example, usually mean you have to be more careful with your authoring and places a greater burden on the artists to make assets that light well with that specific tech. Different technologies are more suitable for different types of games.

        Content creation is a different bag altogether. There is a big push for proceduralism and procedural content creation tools, but the issue is that you generally end up with having some kind of trade-off between quality and breadth, and either way simply creating procedural systems to create assets for your specific project takes a lot of development effort.

      • 9of9

        (2) As an example, you can have a technical artist spend two months making a generic furniture generator and have it spit out some slightly different, blocky, simple couches and tables and chairs and stuff and then use that to populate your world, but they’ll probably look more like something from 2006 and won’t be particularly interesting.

        That technical artist can spend two months and make like a really good chair generator that makes really pretty, AAA-quality chairs in the style of your project.

        Maybe if she spends six months on that chair generator, you’ll have something that you can use across multiple projects and styles, but then you’ll probably spend a month or so just defining the parameters so it creates the right kinds of chairs for you when you start a new project that needs chairs.

        Or like, you can have an artist sit down for a month and make six chairs for you, which is probably all you’ll really need without players ever noticing there’s only six chairs in the game.

      • 9of9

        (3) Using evolution and genetics for arbitrary assets would be awesome, but that tech is nowhere near where it would be usable for production yet.

        So it’s like, all of these things are doable, but you really have to pick your battles. No Man’s Sky probably pushes the envelope as much as anyone does in terms of procedural generation, but obviously the creatures it creates aren’t really assets you could use in Gears of War, and they look fairly samey after a while on top of that.

        On the other hand, Ghost Recon Wildlands uses substance for semi-procedural texturing on a lot of its assets and that saves some work, so you can have a hundred different types of camouflage with their own wear-and-tear patterns applied to your guns.

        • Sam Illingworth

          The procedural work CIG are doing on Star Citizen looks very promising. They have a tool where you paint biomes onto a planet and then the procedures fill them in, so you don’t end up with Star Wars style single biome planets like in No Mans Sky. Now obviously Star Citizen is a very expensive game, and we haven’t seen ho well this tech will work in a finished product yet, but it bodes well for the future.

          Also, reusing generic engines and art is surely a good way to make games cheaply? Isn’t that what Vanishing Realms uses? And that’s great.

          • 9of9

            That’s not really the same thing. AFAIK Star Citizen doesn’t use procedural assets. Painting biomes is a case of distributing assets across the terrain in a way that looks organic. Obviously that’s one part of the procedural puzzle, but it’s not really what @trekkie was talking about.

            Star Citizen has a fairly tightly curated world at the end of the day. They have X amount of planets with outposts and things set down on them by artists – and then generic wilderness covering the rest of the planet. Procedural placement tools are super useful for that kind of thing, but their assets are generally hand-made, and those biomes are are defined by the placement artist (this biome is composed of this set of rocks, this set of palm trees, this set of flowers, with certain rules and parameters for how they’re distributed).

            The main difficulty in going from like, one-biome planets to multi-biome planets is going to be memory, I imagine. If you know that you’re going to keep this one set of assets for as long as you’re on this planet, then you can plan around that and you can use the time spent taking off and landing to stream/generate those assets and fill up the memory with a nice set of variety. If you have multiple biomes, its harder to predict when you’re going to transition from one biome to another, so you’re more likely going to need to store all of the assets in the memory at once, which would mean less individual variation within a given biome.

    • I’ve made several games by myself, thanks to stuff like the Unreal 4 Engine, and all it cost me is my time. If you want to make a game, just make it. There are a crazy amount of tools to help, and YouTube guides up the wazoo to walk you through ever step of using them. It’s quite amazing what’s available now.

      The biggest issue with indie games isn’t the “making” of the game but the “Marketing” of the game. It’s so easy to make games now, even very good looking ones, that the market is flooded and very few games ever really become commercial successes. It’s hard to make your beautiful snowflake of a game standout in the blizzard of games out right now.

      • trekkie

        Walter, the context of this comment stream is the Robo Recall class games. You cant make such a game yourself. You can however make tiny 2D games or 3D games with a couple of boxes bouncing around but if you need characters, a believable immersive environment for VR then you currently you need a team and many millions. I was addressing this issue.

        • Trekkie, my statement stands, and also I think you are very ignorant of the quality of indie games available today. Your comment is insulting to the very talented people who make quality indie games. Most of the VR games available now were made by small studio, independent developers, some of which have VERY high production values. They are certainly not “a couple of boxes bouncing”.

  • Justos

    Theres also the fact that epic makes no $$ from people buying it. Oculus paid for it to be free for Touch owners. I don’t think it was ever 10million to “build”.

    • benz145

      This was the root of the misunderstanding I think. People took some headlines and ran with the notion that ‘Amount Oculus may have funded Epic to make the game == How much epic paid in development costs’. Since the game was launched for free, Epic needed some incentive beyond break even to use their time and talent to make the game. That may be the origin of the $10 million number, but it’s not likely what it cost to develop the game.

  • Pedro Kayatt

    I could do it with less than a million :D

  • Ok, now I’ve just to find 4Millions and start developing a great VR game

  • Fanatoli Guyoff

    WTF I’ve worked on 4 major games and 3 indie games as a mid level programmer who knows most programming languages inside and out……….and I’ve never had a 150k salary. That’s pretty generous

    • Richard Wall

      You also have to remember the inflation of living expenses in certain areas which significantly bumps up wages and averages artificially

      • Fanatoli Guyoff

        Yeah I guess living in the bay area or New York might provide better wages than it does here in Los Angeles.

      • benz145

        Good point, it should be mentioned that Epic is based in NC, rather than some of the expensive game industry hubs like Seattle, SF, or LA.

    • No Spam

      Two important points:

      1. The point of an article like this is to provide generous estimates, to give an upper bound to development costs. If the upper bound, even with generous figures, is still 50% of the commonly thrown-around figure, then it gives the conclusion (that $10m is too high) more legitimacy, not less.

      2. You could also think of this as a “total compensation” number that also includes cost of healthcare and other ancillary employee benefits (like shipping/performance bonuses, 7.5% social security tax, etc.). Often “fully loaded” compensation is 50% to 75% higher than the employee’s take-home salary. So think of it as $100K of salary, $50K of other costs the employer absorbs.

      Does that make you feel better? :-)

      If not, maybe you should submit your resume to Epic!

    • Full Name

      Well, it also needs to include health insurance, taxes etc

    • NooYawker

      I noticed a huge uptick in developers from India working on the cheap. So that affects what you get paid.

      • yasar

        I really don’t think that Indians are working on the cheap affects. As many Indians have no idea about VR, they know about google cardboard (cheaper stuff) though. The cost of Graphic cards are so high due to the import duties and Vive and Rift are not yet available in India. I dont think they will be many who will be attracted towards Gaming VR until 2018.

  • NooYawker

    It’s a fun polished game but no story, and only a few levels and enemies. A few million seems about right.

  • Lucidfer

    So given that this must have taken about 4-6 months to develop at part-time from a few employees, you can reduce that to a 1.5 millions top, and that’s a stretch.

  • fuyou2

    Just a marketing trick to make it look bigger than what it really is. Btw, What ever they spent it’s a piece of shit game..

    • Texazzpete

      It’s one of the best VR games out there.
      Your post history, though. Such childish comments…

      • yag

        Please don’t feed the trolls, you can block them via disqus.

      • fuyou2

        One of the best ha?? It’s fucking BOOOOOOORING AS BATSHIT!! Runs like shit, just eye candy.

      • SandmaN

        Agreed. This kid’s vocabulary consists mostly of profanity and negativity, for no good reason from what I can surmise. As one who is on this person’s equivalent emotional and apparent intelligence level would say, sad!

  • yag

    Wave shooters are not expensive to make, nothing new.