Porting a Real Life Castle into Your Game When You're Broke

[Music] hello so how's everyone doing today oh that's good ok so first can you please put your phone on silent and the quick reminders is reviews at the end of session please fill them out because we've let us know how I did ok great so thank you for showing up for my talk putting a real life castle into your game when you're broke I'm Joseph and my technical designer at words void so a little bit of background word void is an open word a puzzle exploration game and it focuses on historical location of the world and we try to incorporate the rich history of those location into our game so I was working on the bingo site and if you're not familiar with that site it's the home of the Phoenician one of the most Asian cultures its consisted of traders and it's believed that they invented the alphabet so the city itself is located in Lebanon and it's over ten thousand years old so what's interesting about the location is that multiple civilization went through it even Greek

the Roman and the Crusades and speaking of the crusade that's when the castle was built using Roman ruins now the castle itself is fairly recent and it's just 800 years old and I had to scan it for my game so as I was promised the funding that kept being delayed and I decided just go out and do it tools whatever cooking I can find and it worked so this talk is a reflection about why it worked and what I used and hopefully it will help you scan your own castle onto your game ok so the first approach that I took was to draw from reference so I had a few photos of the castle and I tried to draw it and I started to get really decent results though the problem is it wasn't until a few months later that I realized some dimension were off like for example there was a bridge for the castle and in the game it barely fits one character and in real life it also barely fits one character and his horse and his personal marching band so that was nice okay so that's why I decided to just

measure the castle and so I took a measuring tape and I went there and measured every single Rock of the castle took me around two days nine packs of Oreos and I did it basically so yeah okay luckily on the last day I just remembered there was an app called 1 2 3 D catch and that app was just to grab your phone take a few photos of an object and you upload it to the cloud it will generate a 3d model for you so I was like sure I can take a quick screen or a quick spin around the castle so I did that and I used recap and I got this blob so this blob is actually really interesting because even though I cannot use it in my game it showed me that 3d scanning is probably a good way to actually approach this so I started doing some research and I found out that it's named photogrammetry and well I I got to read that in depth into the subject so to be clear this topic and this talk will not cover photogrammetry it's not its reduction to photogrammetry though if you're interested about

knowing more I wrote an article called everything you need to know about photogrammetry I hope and you can check it out though if you're not familiar with the topic it's really simple and just taking multiple images of the same model and if you have enough overlap between your photos the software will generate a model for you so I starting doing some research and joined some group and I found the Facebook's 3d scanning group which I highly recommend you join and it's filled of industry veterans that's more than happy to share their experience and on that note special shout out to Jeffrey n Wilson who actually helped me figure out this list of equipment that I needed for my project and first I need the DSLR some accessories a drone and a special rig and this rig is used usually for capturing the environment in order to remove it later in post process so let's see how we can actually reduce the cost of every single one of those equipment first let's check out the camera and since the camera is wrong what a decent one for photogrametry can cost up to 3k

there's actually you don't need a very expensive camera and this is a living proof and this game is done by me loves luka and he's an the member of the community in 3d scanning Facebook group and what's impressive about the scan is not the level of detail he achieved but it's actually that he used a camera at the SLR that's in the Canon 550d which is not one of the highest end cameras he also used an adjustable lens and to compensate for the lack of resolution what he did he took more photos he actually took 436 photos of a single door so and he was really captured so you don't need any expensive DSLR and most of them will just work some people even use their phone for example just save in the high settings or save and run impossible if you can buy it at rent or borrow and also you can always post process your photos which will help you get better results and will boost the quality so I used the X objects you can use Camera Raw or Lightroom 2 so as for the accessories so fixed lens will

probably give you chopper result but you can use an adjustable one just make sure you pick one focal length and stick to it do an entire session of the capture and SD card because you're gonna need a lot of photos and this one is actually obvious but well before my battery cannon ever drained and twasn't I just want to scan the castle when it did so always get an extra battery luckily security guard over there was kind enough for me to let me charge my battery but I did lose a couple of hours of course you need a tripod and a monopod for more stable images and the remote what $2 so get it anyway and basically for the remote well basically when the camera cannot get to something when you want to take photos of some areas that you can't really reach the camera you can use the remote and it's really useful as for the drone it will definitely make your life easier though I did not use one and it's because I couldn't afford one the drone I needed was a DJI inspired

cost around three thousand dollars but now you can get similar result by etruscan in the market proaches one thousand dollars so if you can't get one here's how to probably do a scan without it and the first thing you want to do which is just scan the floor and then go one level over it scan the second floor and then take a look from the second floor where both floors are crowing and if you look at the picture at the top right you can see both ledges of the two floors showing which has the software actually allowing those logs together so how do you scan the roof because the roof is there isn't anything on top of it and what you need to do is use ground control points and grout control points are basically points that you specify manually on multiple images it will have the software align it though once you have a point on the roof it's not actually visible from downstairs so how do you actually find a good reference point and in my case I got lucky there was a pole I also don't know why there was a pole but there is one so yeah that's great so look for

those features if there isn't any try to put one and it will help you manually actually align the roof I'll get a drone if you can that's what's a special rig I did not use one because I don't have Amazon in my country and it's really expensive to buy one that's already built so I'm glad I didn't use one because later at GDC there was a talk called photogrammetry and Star Wars Battlefront and they mention that technique was a used object normal space object space normal map and basically it's useful for removing the light without any equipment it will give you similar results so that's great so as for the software since I was using one two three D cash I look for the Autodesk solution and I found something called remake and remake is actually free if you use the cloud and you can upload up to 125 images so if you get the professional version for $60 a months you can upload out 250 images and you can also do some offline processing the remake didn't give me a lot of controls on the process itself and basically that

may or may not be something that you want in your project and so I look for the other software which is the most commonly used back then and it's called ESRI soft and azure soft it's only a $190 for a license and has unlimited amount of photos that can handle though actually underneath it was an Asterix because the more photos you have the more RAM you need so if you're gonna scan a castle and imagine you need around four thousand six hundred or six thousand photos you're gonna need probably 512 gigs of ram and that's gonna be a very expensive server if you want to use a ground control point you need the professional version so that's an extra $3500 so solution cannot be can be quite expensive and that's when I looked for reality capture and rewrite capture is 99 euros per month and it's actually a really fast much faster than reality capture and what's interesting about it it can run on almost any laptop or desktop as long as it's to have an NVIDIA GPU and can handle an limited

amount of photos no matter the random amount of RAM you have the professional version sorry the promo version will probably just handle up to 2,500 images while the yearly license will hunt it will handle unlimited amount of images so if you're scanning multiple castles during the year I would highly recommend you invest in a yearly license so if it's a one-time project then you can also consider pics for D which is $350 for one months but it can also give you a similar result and also all of the software don't give you same result out of the box if you invest your time in every one of them you can actually get the same result just by tweaking some settings so as for the PC for the PC were according to my testing the other the recommended specs that I found out GPU well you can use an AMD but if you want to use the edge capture you need nvidia so Nvidia is a safe bet in this case I had some issues with match to match lab for vram that's less than four games so get one with four games for the CPU you might need a seven or XE on and

in order to make that decision know that I seven is limited to 64 gigs of ram the reality capture for example prefer faster cores over more number of course while edges off prefer more cores because you have to disable some of them if you have more GPUs and more RAM so it depends as for the RAM amount found 32 gigs to be a safe bet I did find some issues with 16 but most scan will actually work with 16 unless your scan in castles and this number won't actually really help you energy soft but others also who actually works great as from the hard drive well you can get an SSD but it's really expensive so just get one that has small capacity to run the software on it and you can store your photo on an HDD and the thing with the HDD is that it's not expensive and it will just probably add some time to the first processing layer which is loading the images into memory okay and please don't use a laptop I mean I totally ignored my own

advice and I use my own laptop actually because I was broke but you shouldn't I mean okay also most of the advice I give actually broke all of them but you shouldn't great so I mentioned that I went scanning the castle and it worked but actually there was a long period of pre-planning and practicing before that so if you had ever scanned before start by practicing it for example for me it was the rock in my living room that I start skating and took a few images try to take less take more see what happens and yeah then you can ask yourself how much resolution do I need do I really need all the information for example if the wall is offset by 10 centimeters are you an archaeologist so you need a huge amount of precision or are you a game developer and you don't really care do you need all the details already covered in the textures and those questions will actually define how much force you will take you can probably take six photos for one wall or fifty photos from one Rock okay so also you don't have to scan

everything in high resolution just look for some features like the textures if you want to texture focus on that some modular pieces and even in that castle where nothing is modular because it's really old and every pieces is not identical arrow and nothing is symmetrical I still found some pieces that are coming to the architecture of that time and I focus on those so for example the stairs the window or the ceiling okay so for the pre-planning then you have to imagine how are you going to capture your photos so first you want to take a huge loop around the castle so I'm drawing the loop right now it doesn't update in real time okay and this is the first loop we'll just cover all the details of the castle and make sure everything is all the photos though it's not enough you have to take a second loop that's closer and covers more details so one second okay and that's the closer loop the thing is there's a hill over here so

you can't really take photo from here and you have to climb all the way to this point though the problem there are some spikes so that was challenging and covered most of my field of view there's also a bridge on there nice over there and some typed messages that you have to also compensate for and also not to mention the Sun is not your friend even if you're shooting in a cloudy day a small Sun shifting can actually ruin your entire scan so the better way to approach this is not taking a look because they can look and take up to 10 minutes and in 10 minutes the Sun can change a lot so instead you wanna scan each element individually so example you scan the first hour then you scan the wall then use kind of laser tower and you sort it into folders because just in case that happened you can always just manually align those elements but if you took a loop and doesn't close then it doesn't gonna work out really well actually for you so that's something to consider so as for the interior and this

also works on exterior don't take panoramic images and parameters are basically images that you take in place so don't do that walk while you capture the photos because this will generate a lot of errors and I took many panoramic images so just knowing okay so when you want to scan some tunnels it's really hard to move inside of tunnels so make sure to grab a photo and then take one step forward grab another photo and keep moving on forward though if it's a stairs for example direction actually matters because you want to go up the stairs and not down because you want to see the steps in photos another issue that you might face is dark tunnels and in this case some areas can be much darker than the other and when capturing with the petroi pod you can probably change the shutter speed but that can backfire too so you have to be careful with this one but it will help you get more dynamic range or more points to actually align the images so speaking of alignment

once I captured all the photos I put everything in Riyadh to capture and hit a line and I got a big blob so I the problem is I didn't actually take a screenshot of that blob and I want to try to recreate this process just to take a screenshot of it I got a good result okay so I know it's also actually improves this time and that will get me to one point which is whatever I mention right now was sticking a year ago so you have to do your own research because they change a lot okay though there are still some errors like if you can check out the red areas over there you can notice that some walls are magically appearing somehow and a way to fix it as you split it into components so reality capture has a nice feature when you can align some group a group of photos create an element and then align that element to another element and that way you can see which photo actually is not working remove it and reduce the error and of course you can save them an organization well yeah use a naming convention not like I did okay so and

then whatever doesn't align you can always use some ground control points and align them and you can end up with this and what's interesting about this is that see those points this is actually me standing there and taking a photo so that's around four thousand four oh that's nice okay and this is the end result and then you create a Pollock launch version and you put it in game engine and now your gameplay programmer can actually work on the gameplay while your artist can slowly start replacing the pieces with more better looking one and if someone was taking not during the presentation yeah I'm sorry I actually took notes for you and I will written in an article so you can check it out Scott scanning a castle inside out with no budget okay so how much will it cost you so if i want to scan a castle today problem that I need drones $1,000 a camera $300 camera accessories a PC rillette capture and a Photoshop or DX

optics and that will give you a total of 12 K how much it cost me 70 bucks and the thing is I do not use a drone I borrowed the camera okay a camera accessory I didn't actually buy all of them the PC actually won my laptop and also huge shout-out to Melo's who actually volunteered to process the data for me on his server so I didn't actually have a process server reality capture was an open beta when I tried it so I got it for free for commercial use and the X optics I use the trial so that's how we scare reduce the cost and it's about finding opportunities and how we find opportunities well if you're at GDC that's a good step you can also join a 3d scanning Facebook group ask question over there go to events make friends and for example if the software doesn't really does it for you you can just try to contact the developers and try to work something out with them and always stay up to date like for example if you're following DX optics you would know that they released the previous version for free for the next few days

so that's useful and ok great you scan the castle now what and the thing with castles is that it's 1 billion polygons so good luck finding is also that can open it so the first step is to probably take a to Polly crunch it to 15 million in reality capture because like I mentioned you don't need that much resolution second step remove all the details it's really distracting and I'm it's really tempting to keep everything but you have to remove it because you're gonna become really over run really fast and then split the castle into pieces and then create a schedule and for example if let's say you need three days to create one piece give it five days just in case otherwise you really gonna become overwhelmed though you still end up with two million million polygons piece and you still have to work with it and I struggled a lot with that process because I couldn't find the software that can handle this huge amount and the thing is you don't have to handle the huge amount the

operat method that we still use still work really well just need to find a way to translate them into the new polygon an example for example you want to UV a model that's 1 million polygons and probably 3 max from Y I cannot open the software ZBrush may or may not actually work and the best way to do it has just reduce it into a lower polygon UV that and subdivide and we project the details and that's actually gonna work really well because the small distortion are not really visible in texture so you can get away with it and speaking of old methods you can always block your model and remove even more details because they are still distracting and you don't need the any of those and this will help you actually figure out the volumes and which pieces you want them to become modular and yeah so if this is the first time doing that I would highly recommend you perfect the process so imaginary topology can take up to two hours to finish okay so maybe invest a week or two but understand the process perfect it because that's the only way you're

gonna become better at it and it's gonna become really fast in time so don't be perfectionist because no one's gonna notice basically that's it okay also invested in your textures so my castle is made out of rocks so if I spend one month creating this really nice procedural generated shader as rocks then that's great it's okay I can budget that and if you're interested about the process of doing that there's also another article you can check out and it's called the workflow of creating game ready textures and assets using photogrammetry okay so once you become really comfortable that the workflow you can actually do some interesting stuff like they take this european statue separate the properties of the statue from the properties of the material and then do some transmutation work and get a Middle Eastern statue and if you're interested in knowing how I did that actually wrote an article but it's not posted yet so you have to follow me basically in order to get it sorry okay and here's a long list

of free software's that you can check out and my advice and basically just do it because it took me around six months to scan the castle now in my defense I had to learn everything from scratch but could probably have done it sooner if I just went out and do it from the first place and if you ever actually it's kind of castle or do something else remember take some time to go back to the community and thank them because probably someone helped you and speaking of that a big shout out to my mentor nada the two industry veterans Minos and Jeffry our archaeologist Hassan and my aunt who let me borrow her camera Samia I will also recommend that you join this reef at a scanning Facebook group and a special shout out to the American University of Beirut send me to Austria to see a 3d scanning conference so if you have any questions you can catch me in a wrap-up room after the talk and the no slower Lobby by the elevators you can check out all the articles by going to my website word void and checking out the developed section if you have any question you can post them underneath the blog post and you can just ask me on twitter directly

and i'll be more than happy to guide you and set you on the right path okay so yeah I'm done thank you so any questions so for each apologizing are using topogun or will help what program okay so actually using three softwares because apparently when you have 3d scanning you still kind of find a match it has some issues so sometimes I lose a use remake I Polly crunch your model or sometimes I use 3d max and I also click on your model I also use instant measures sometimes and there was a polychrome shoe model into some elite apology and generate a quad match and apprentice for and I use also ZBrush sometimes and ZBrush has to process dynamic and the receiving measures Ramesh yeah right so for zero measure

you can just do some little policy but it can become a real low resolution model then you have to subdivide it and then we project the details from the high resolution model and that's how you get it alright cool thing oh and also feel free to use a combination okay I just wanted to say if you don't have anyone you can borrow cameras and lenses from there are plenty of places on the internet that will rent high-end cameras for fairly cheap and on the order of like tens to $100 a day so if you need drones and cameras that might be an option as well all right my anti the camera phones like hey can I borrow your camera yeah just saying if you don't have an aunt you know yeah okay cool any more questions okay great okay just don't forget to fill out the reviews after the talk and I hope you enjoyed everything thank you [Applause]