thepixelcrush

[thepixelcrush]

A Treatise on McQuevian Hand-Painted Texturing Automation

Originally posted on the Big Robot blog, Gamasutra, Kotaku, and RockPaperShotgun.

What follows is an expanded take on what we touch on in that video. If that sounds good, read on!

Tölva’s world and tone are exciting for a number of reasons – lasers, exploding robots, unfathomable space mystery and the phrase “combat archaeology”, to mention but a few – but here we’re going to talk about how we crafted that rich, screenshot-friendly visual style. We will try and understand the majesty of Mr Ian McQue’s concept art, learn the fundamentals of ‘splattery precision’ and some other made up terminology, as well as comprehending the beauty of a clean normal map bake or efficient UV layout. You know you’re in deep when you can appreciate UVs. We intend to go deep(ish).

So… Defining the look of Tölva was a process of looking at our inspirations and influences and breaking them down into their constituent parts. This way we could take what made a certain look ‘work’, and begin to create a process that we could apply to a variety of assets while maintaining consistency with one another as well as looking appealing in isolation. This meant taking Ian McQue’s sketches and paintings and finding common shapes, colours, types of brush stroke, silhouettes, and starting to form a “McQuevian language”.

Often robots, vehicles, or ships, were chunky with lots of rectangular protrusions breaking up their silhouettes. Layering of details was important, with colour often going from darker to lighter as the layers got closer to the exterior. Angles were often a few degrees off being parallel with each other but still consisting of straight lines, rarely curved. The dumpier a spaceship, the thinner it’s aerials and wires: contrast from form.Often robots, vehicles, or ships, were chunky with lots of rectangular protrusions breaking up their silhouettes. Layering of details was important, with colour often going from darker to lighter as the layers got closer to the exterior. Angles were often a few degrees off being parallel with each other but still consisting of straight lines, rarely curved. The dumpier a spaceship, the thinner it’s aerials and wires: contrast from form.

We could use these style guidelines to judge how well something would fit in the world, and how close it was to the source. I’d sometimes create reference boards using a tool called Kuadro, a convenient way of laying out image files on the desktop and storing their scale and position in a file. I’d be conferring with Jim at various stages throughout an asset’s creation- he would direct the design, and mine his extensive and tasteful tumblr image library, or doodle thumbnails to illustrate specific requirements. Interpreting sketches from one angle and turning them into 3D geometry has been a particular challenge for me on this project, but following these guidelines, and practicing this approach has enabled me to get better at it through the course of the project.

The first step towards actually making something, once reference was set up, was blocking out a 3D form in Maya using simple primitives and keeping things low detail. Often if this stage went well the geometry could be reused for the final low polygon asset. Then I’d take this base mesh into zBrush and play with the shapes, maybe add things that were more awkward to create in Maya. If I was feeling really fancy, I might do a paintover.

Sculpting, Detailing, Baking

Once something is starting to feel like a strong design I can start to commit to adding details, again referring to the guidelines as I bevel edges, chamfer corners, and smooth any (rare) curved surfaces. Then come surface details that will be baked down into the normal map such as wires, nuts and bolts, screws, handles, cutlines. It’s easy and enjoyable to go overboard at this point and fill every surface with greebly goodness but any artist will tell you that the eye needs space to rest. Not only are you looking for space to rest within this asset, but often you will need to look at the game as a whole and decide whole assets need to act as rest points devoid of any noisy detail, in making up the composition of the world.

At this stage I’m also starting to break up edges with wear and tear and damage where seems appropriate, this is one of the most satisfying parts- chipping away and scratching up clean surfaces requires little moment to moment decision making and I can let this almost therapeutic activity absorb my attention. It’s important to split objects as I go into logical groupings based on their material types, I will use these later to bake a material ID map which is essential to the texturing process.

One of the of the recurring aspects of creating such a mechanical world is designing (relatively) convincing joints, paneling, and other robotic goodness. It was important to spend enough time look at machined parts or engines so that I could start to internalise where a cut line might occur, which panels needed screws, how a curved edge might be carved out of a larger piece. Manufacturing is often about starting with a form and reducing it from there, and the same approach works really well when sculpting.

Then comes a series of steps that are dull but very necessary to getting a clean bake but once that’s done I can move onto texturing. A tight UV layout, good smoothing groups on your low polygon model, and a precise cage mesh all aid in a clean bake.

Generating an Automated McQue Texture process Or: How to Hand Paint Textures, Without Using Your Hands.

Good modelling is essential and a stunning texture won’t save a bad model, but texturing drives a lot of Tölva’s look and helps distinguish it from other games. It’s also what was most clearly imitable from the concepts. We were fortunate enough to be given access to the list of secret brushes that Mr McQue utilises most frequently and to such effect, and using these able to create an automated process that streamlined the texturing of hundreds of assets. We essentially created a system where I fed in my sculpts and a McQuevian texture was churned out for me to tweak and finesse. This worked by taking the right brushes in Photoshop and creating a tiling pattern with each, attempting to recreate a certain type of stroke or scribble that could be found in any given concept piece. With these ‘master strokes’ I could tile relatively convincing McQuvian patterns across a surface, tinting its colour, and using the pattern to break up the edges of masks in Photoshop. Daub, dash, splatter, speckle, spray, smear, squiggle, strokes, crosshatch, and grainy swirl would become my best friends over the next 2 years. Grainy swirl was great for organic or noisy surfaces like mud or concrete, while crosshatch had a natural galvanized metal flakes look. Each pattern came into its own as I built a library of reusable materials from these tiling patterns.

The tool tying all this together was the Quixel Suite’s dDo, a Photoshop based texturing tool aimed at primarily at creating PBR materials very quickly, which we had employed for our own nefarious aims. Feeding the sculpt data from the bake (tangent and object normals, ambient occlusion, material ID, gradient, sometimes a height map) dDo could tell me where edges were, crevices, what was at the top of the mesh, what was only facing down or upwards. The amount granularity in terms of how you define the masking of a material is very powerful, and once I’ve made decisions about where a material should be confined to I can apply that as a preset to any other asset. “Here’s my paintedMetalC material, it will have chips and scratches on the edges, sunbleach on the top, water damage on the undersides, and some lichen in the crevices.”

I found the simplest way to structure a texture was to have a form layer at the top that brightened upward facing surfaces, and conversely darkened downward facing surfaces, with an edge brightening layer to accentuate the objects natural shape. Our ambient light in-engine is a single colour and so baking in some subtle lighting data helped objects have some shape even when in total shadow. Next, a weathering layer that universally affects the asset. Things like dust, stains, and other environmental effects go here. Beneath this all the materials are applied to their corresponding material IDs from the high polygon sculpt bake.

From this point I’ll start to work in overlaid scanned details from the dDo library, these won’t affect the diffuse channel very much (deliberately), but will add some physically based detail and reflectance values. The materials that make up Tölva and it’s inhabitants are 90% diffuse covered and 10% specular, meaning they are largely matte looking surfaces with bits chipped away to reveal the metal beneath. McQue’s style has very little exposed steel, or polished chrome and there is no glossy plastic, so I tried hard to match that. More on the PBR side of things later.

Once materials are assigned to the whole texture there’ll be a lot of tweaking of the tiling pattern’s intensity, colour, size. Material specific details will also go on, like rust that only affects metals. Then in some cases I’ll add bespoke hand painted details like glyphs, diagrams, or bits of weathering that only appear where there’s a pipe or something and that can’t be defined by the automated process. But this is the only part of texturing where I’ll actually paint a specific detail onto a specific part of the texture, everything else is controlled by dDo and my masking.

The colours we use are often collections analogous colours making up 60% of an asset, then it might have a darker or more saturated variant covering another 30%, and then remaining 10% has a bright or rich accent colour. Usually the best way to get a decent start with this is to colour pick directly from a concept and go from there, can’t get more accurate than that. As we got further into the project certain palettes would mean different things to me about where they were in the world and who created them, making decisions about colour schemes much easier.

Once all this is done I can save individual materials as presets, or entire documents to be reused on similar assets, this is largely what facilitated us in getting the texturing done as a efficiently as possible.

Implementation: Make Everything Modular

Being able to cannibalise existing assets for reuse is an invaluable tool when you’re trying to squeeze as much variation into the world as possible without burdening your VRAM usage further. The mileage you can get from an asset is often surprising in terms of reimagining it for other purposes. If you model things with discrete watertight intersecting meshes rather than combining everything and deleting interior faces, you can at any point split these out and reassemble them into a new variation of that asset.

This is probably obvious to veteran game artists but was something I was often in two minds about at the start of production, modularity vs bespoke detail. The other thing to bear in mind with this approach is baking with your meshes well exploded so occlusion data doesn’t interfere with neighbouring meshes, preventing you from reusing that one piece in a prominent place because it has a massive shadow across one side of it. Meshes can be merged using boolean operations to merge intersecting meshes and create totally new shapes, which we did to create cliff faces comprising multiple rotated cliffs into one uber cliff.

Creating asset kits is a common technique in games utilised particularly well by the Bethesda teams. Tölva has less of a need for complicated interconnecting architecture but we were still able to get some use out of concrete piece kits that we made to build more industrial areas, and kits of damaged spaceship to decorate the space wreck debris that litters the world. Creating these kits for the environment team (er, Jim) took a sizeable amount of the art dev time. Another example of this kit-based approach are the weapons attachments, all modelled as separate items that could be rearranged and mixed together to create weapon variants.

Working in the sizeable world of Tölva meant pushing a lot of geometry onto the screen at any one time. Occlusion culling and culling regions far from the player helped alleviate some of this but ultimately we were going to have to create level of detail meshes for a lot of polygon heavy, or frequently used assets. This is a fairly simple process and its amazing what you can get away with when an object is at distance.

Expanding the look with Shaders

I have no programming skill or experience to speak of, but I found fairly quickly I was being limited by my ability to create simple shaders to achieve effects commonplace in games and desirable within our sci-fi aesthetic. Unity plugin Shader Forge became my go-to tool for solving these problems, it’s a node based shader writing solution that is approachable but still crammed with maths terminology most artists have never heard of. Using this I was able to assemble and visualise shaders that had pulsing lights, tinted a material based on ID maps, did weird things with transparency, or just layered textures based on normal direction. This saved me using up precious programmer time – Tom was full time on the project, and Dan part-time, so their brain cycles were at a premium! – and gave me control over very specific parts of the look. This kind of autonomy is very precious when you work remotely and there isn’t always someone on hand for you to pester with technical queries. Being a visual scripting system it does have the downside that when it breaks, I often had no idea why, or how to fix it. These issues weren’t insurmountable but did lead to a day of debugging from our resident shader expert, Tom Betts.

I would approach materials very differently if I was starting this project from scratch tomorrow. Coming from a CG background pretty much everything is given a specific texture that uses the full UV tile, and that was pretty much my only option to get the look we wanted at the time. As my understanding of Shader Forge evolved I was able to create a sort of shader version of how my textures were set up inside dDo. I was limited to 4 masks (RGBA channels of the ID texture) plus a base layer, and I could tile and tint some of the McQue patterns within those masks. Each McQue pattern had the diffuse pattern stored in the red channel, specular in the green, and gloss in the blue.This allowed me to get good resolution on textures for assets that were 5, 10, 100 metres long. The ID and normal maps were still limited by resolution but the tiling worked very well. It serves well as a halfway house between a 1:1 mapped texture and tiling materials, but lacks the subtlety and detail of a 1:1 or the variation and fidelity of having a library of materials applied to submeshes.

Terrain texturing

Our initial approach to terrain textures was to treat them as purely two dimensional, abstract patterns. As the style developed it became clear we were going to need something a bit fancier to fit the aesthetic, especially given the fact that terrain filled the majority of the frame a lot of the time. I began sculpting tiling terrain materials in zBrush, creating quick sculpts that could be exported to Unity quickly to test tiling, detail, and blending between layers. It was important to use detail meshes like small pebbles and shards of rock that already existed to tie the textures to actual 3D props sitting on the terrain. This was the first organic thing I’d had to sculpt that wasn’t a rock or cliff and the soft shapes weren’t fitting the style at all. The solution ended up being to reduce the polygon count of the sculpt using zBrush’s decimation tool to create a more faceted organic surface. Feeding this through dDo and using existing material presets was pretty straight forward. The height map was used by the RTP terrain setup to cleverly blend layers together based on depth, creating a much more pleasing blend that a simple fade.

Lighting/Rendering/Post Processing

PBR or physically based rendering has become a buzzword and marketing tool as it’s adoption has spread, symbolising a graphical upgrade over traditional game rendering. Mwoar detail mwoar realism mwoar immersion. Much as I love all those things it’s most important contribution from the artist’s perspective is having a controlled environment within which to work where you can create a material and have it react to light in a predictable manner, not only between scenes and lighting setups in the game engine, but also between software packages.

Having this predictability is even more useful when your game has a full day night cycle and is going to appear in all sorts of lighting conditions. Even if your game isn’t crammed with brushed metal and lacquered pine (highly reflective surfaces being a sure sign of a game trying very hard to get the most of its PBR) it can still massively benefit from a physical approach. Having a nice fresnel falloff on a largely diffuse material makes every object lit at a glancing angle look fantastic, materials are energy conserving so nothing washes out in a way that just appears like a white value being clamped.

McQue’s paintings have naturalistic quality to their lighting, this is ideal for PBR. His colour palettes have a fairly limited range, there are few bright whites or really dark blacks, much like real world albedo colour values. This allows a PBR lighting model to really shine without being overpowered or limited by overly saturated or contrasting texture colour values. PBR is not a style, it’s a tool. You can push your look in a number of directions with colour correction, lighting, and post processing- but you are starting from a well calibrated, stable, neutral position that is easy to control. Using PBR also means you can draw on nearly 200 years of photographic knowledge. Exposure, tone-mapping, HDR, colour correction, bloom, lens flare, chromatic aberration, depth of field, vignette: these are all techniques that have either been part of photography and film for a long time, or are simulating things that a lens does naturally. Also use linear colour space everyone, it has been VFX industry standard for forever, there’s no reason not to.

Having this linear colour space PBR setup means we are generating accurate colour values in the render that, after tonemapping, exceed the screen’s full brightness value. They are brighter than white, this is what makes the render high dynamic range. We use these values on our light emitting (emissive) materials particularly to trigger a bloom effect, but also to illuminate any dirt or scratches on the lens you’re viewing the game through.

Nothing says sci-fi like bright blinking lights!

[thepixelcrush]

The Signal From Tölva

It’s been 6 months since I last posted on this blog, but I finally have something concrete I can share.

In January 2015 I left Aardman and embarked on a game type adventure with the creative people at Big Robot, ever since then we’ve been chiselling away at a science fiction, open world, exploration/combat project call The Signal From Tölva. Find out more from the team here.

I've been in charge of:

the majority of the modelling

the majority of the texturing

assessing 3rd party models

defining the lighting and post processing

establishing a style in keeping with Ian McQue's handsome concept work

creating shaders to achieve certain looks and effects

some (very) occasional rigging

I have absorbed as much in the last year as I did when I was first starting out in my animation career, being the sole full-time artist on a team means finding solutions to visual problems without a team of experienced artists to draw from. The Big Robot team have helped me transition from rendering to baking, subdivision to low poly modelling, displacement maps to normal maps and now we can finally talk about it and share what we've been making together! Expect more posts coming in the near future talking about some of the things I've learned, and showing some of the shiny things we've created. Exciting times indeed.

[thepixelcrush]

Molten Pixels

As a video game making person I often have a need to make myself a tool in order to build a certain thing, create a certain type of effect, or just complete the task at hand in a timely fashion. Currently, the thing that seems to repeatedly facilitate that is a plugin for unity called Shader Forge. Shader Forge allows me to write shaders without knowing any code, I can click and drag to make connections between textures and the plugin's many different nodes to control how they affect what the material on the model is doing.

One problem that I'd been struggling with for a long time on the project I'm working on was to do with large scale props in the world. Anything intended to take up more than 5 to 10 metres of space very quickly started to look blurry as a 2k or even 4k texture was stretched beyond the detail it could realistically display. I saw the revelatory video embedded above about how this kind of problem is handled in the Unreal 4 engine, by creating basic materials like metal or rubber that are then referenced in a "master material" and assigning them to a mesh based on a pair of RGBA ID texture maps. Unfortunately, Unity has no kind of similar material instance function and manually plugging the textures into one material quickly creates problems that stop the shader compiling (you are only allowed a limited number before you hit pragma fragment pixel something errors).

I was, however, able to make a stripped down version that allowed me a base material, coated with 3 additional coat materials, and then a 4th that I usually reserve for a dirt/rust/lichen material. The reason this approach is so powerful- even without all the clever instancing, batching, and extra channels that Unreal 4 supports, is that I can tile every one of the separate materials creating sharp detail at any scale as long as you tile the textures proportionate to the scale of the object. This does have some initial textures created up front for the materials you're going to use but from then on your saving a diffuse and specular texture for every additional asset that utilises this shader and workflow. I start by making the asset normally and assigning IDs at the high poly to low poly baking stage, these IDs are further throughout the texturing process using masking in the Quixel suite.

These IDs control where which parts of the mesh have which tiling textures assigned, in the example above the base layer (no ID) might be a metal, then the red- a coat of paint, the green- a rusted metal, and the blue a tarnished plastic. All of the textures are a mid luminance grey so that they can have their colour tinted, some have accompanying specular/gloss maps. Here is what the node layout looks like for the different textures being layered up and masked in the shader using the channel blend node.

Using some shader magic, specifically a node given the minimalist name "round" I can round the values of the ID map up and down. So, for example, when a pixel in the ID map is between 100% red or 100% green the "round" node will round it to be one or the other, what this does is create a crisp divide between materials on the model and makes ID map resolution much less of an issue. The only place I don't do this is on the dirt ID where I usually want some transparency or feathering on the way the dirt sits on top of the other materials, this has its drawbacks but overall gives me more control to create like dustings of whatever is covering the asset.

I can expose texture slots, colour tint, and tiling controls on the material in Unity so that the colour palette and scale can be changed pretty quickly. It can become hard to tell textures apart without referring to the ID map though when all of them are grey initially.

This master shader has probably been the biggest advantage so far of using Shader Forge but I've also created my own billowing cloth shader using vertex offsets, a blurry and refraction glass effect shader (this one was slightly agonising to set up as it took about 20 nodes to do the work of one line of code). Je recommende.

P.S I was reading an old blog post of mine from 2 years ago today, it reminded me how much I used to enjoy writing this blog, and how much I miss it now. That, coupled and the burning need to bore someone with details of my shading adventures provided the impetus to write this down.

[thepixelcrush]

The Ongoing Examination of Making Things

Last weekend I had one of those chains of experience where one thought or discovery leads to another new and interesting thing. In this case, it started with this video from esteemed video game people at cool ghosts.

My curiosity was peaked, I loved The Stanley Parable so the prospect of experiencing new things interrogated using a similar structure was appealing.

There's a particular scene in The Beginner's Guide (the second game from part of Galactic Cafe) where a character's creative ability is represented by a machine. While it's probably the most literal section of the game it's also the part that resonated most with me due to the kind of question asked of the machine, and of the player. Why isn't the machine working anymore? Should the machine be destroyed, or the work it created? Is the machine making you unhappy?

This lead to conversations about ideas and making things, the stupid filter we (I) have. So having talked that filter into taking a few days off, I picked up a tool I'd wanted to have a go at learning for a while.

Twine is a free tool that allows you to write passages of text and link them together to create whatever you want; it's primarily used for branching stories. There are also simple programming commands you can get Twine to execute: did the player see this passage already? If the player did, show them this text. Did the player click this word? Then display this passage. Things like that. Here are some of the most useful links I found whilst learning basic "if" statements, and a couple of other things.

http://www.auntiepixelante.com/twine/

http://twine2.neocities.org/

http://twinery.org/wiki/twine2:guide

http://twinery.org/forum/discussion/2620/a-tutorial-to-twine-macros-if-set-and-click-for-twine-2-0-harlowe

You can play my Twine thing here. There are a couple of junctions where 1 of a possible 4 random passages of text are displayed depending on what kind of outcome you get from the machine. These are coded to be random, so there's not a whole lot of real choice going on here, but I'm pretty pleased I managed to get some of these code base bits working. Never managed to get anything like this to work before so that's exciting!

You can see some of the structure of my story in the image above. Notice how some text branches out where the player made a choice, and then the story converges again with the main branch. Some passages are floating islands, detached from other passages, these are generally the ones that are displayed randomly within other passages.

It's felt good to be making stuff, not good stuff, or even stuff that I've tried particularly hard on from a design perspective. Just working on something, learning a few genuinely new skills, and having something to show at the end of it. The act of making that thing is so much more valuable to me than the resulting thing itself, I suppose that value distribution will balance out with a lot of practice.

[thepixelcrush]

The bone-propped Olly Skillman Wilson has sculpted one of our own - Wise Old Gat’rk’cil.

You can see the sculpt process here and see more of Olly’s work HERE

[thepixelcrush]

Pixel Pick Pocket

A 2D, rogue-like, platforming heist game called The Swindle came out recently. I've been playing a fair bit of it, and now I'm done. I rarely persevere with games that annoy and frustrate me, but there was a hook to this game that kept me coming back. I want to say I saw the glimmer of a great heist game that allowed me to fulfill daring fantasies of last minute escapes and bountiful rewards- where everything was down to the last second, and every move counted. But I've already played that game and its called Invisible Inc, and it's fantastic. This game is not about that about greed.

I, it turns out, am greedy. Whilst ransacking a level, the fastest way to accrue large amounts of money is to hack computers. Imagine there's maybe 3 or 4 per level and I'm 2 down, I already have nearly £15,000 (a not insignificant sum by any thief's estimate) and I'm going for the 3rd and final computer. Why? I already have plenty of loot to improve my character's abilities, the 3rd computer doesn't look too heavily guarded, but that's not the point. The point is stuff will happen despite your best intentions and you will explode into a cloud of bank notes, guaranteed. But I can't help it, I can see the computer full of money and  it's within reach so I have to try.

Every. God. Damn. Time.

I mean, maybe if the game didn't constantly have me catching on ledges, being attacked through shut doors, falling to my death after a 1m double jump, being stuck next to my own bombs, etc etc, maybe I would get away with my greedy impulses. But the games clunky execution punishes those impulses every time, and it turns out I do not respond well to this at all. The reason I keep coming back is because I know how good it will feel to clean out every computer, safe, and lockbox, wring the level for every ounce of cash it has and then waltz off into the sunset. But that doesn't happen and I am robbed of my clean get away nearly every time.

None so spectacularly as my most recent heist that was deftly enacted to the tiniest detail. I was patient and watchful, planning how to approach each room, how to tackle each obstacle. Then the triangle glitch struck. Just as I was nearing the last computer I tried to shut a door behind me and found that the triangle button assigned to this, and a number of other important actions, was teleporting me down the corridor. Fortunately with no consequence in this case as I had already cleaned out the previous room. I hacked the final computer (with some difficulty as this also requires the use of the triangle button) and then made my get away. Just as I was congratulating myself on the slickest operation any criminal had ever conceived of, I pressed the triangle to access my escape pod and was unwillingly teleported into a room with no exits, 3 hostile guards and an angry raven. I don't even know why there are ravens indoors or what they do. They beat the living hell out of me and took all my cash. I am so done with The Swindle.

Unless I still think I can make that perfect heist?

[thepixelcrush]

Reel 2015

Having collected a decent amount of footage from the things I've worked on over the last two years I was high time I put together a new showreel. What's nice about this one is there are a couple of personal projects in there that made the cut along with some of the commercial stuff, also some favourites return. Who doesn't want to see Leonard's wrinkly face one more time, after all?

I was just going to use the song uncut, but as someone who often doesn't focus on lyrics I had to have it pointed out to me during an invaluable feedback session that the words were probably a tad too dark/ponderous for what is ultimately a piece of promotional material for my work. Though I had forgotten how much I love editing video cuts to the beat of a song, there's something satisfying about the rhythmic union of image and sound. It works better with some pieces of music more than others, so that's a big factor when choosing the soundtrack.

I also simplified some of the text overlays to try and make things a bit cleaner, there's a lot going on with shots mixed up and chopped about so the easier it is to see what I'm responsible for in each shot the better.

Unfortunately, there's no game stuff in there really. I need to correct that for the site in general. I do have some HeroSquad and Championsheeps material stashed away but not of it is particularly presentable. Mostly lots of sprite renders and work in progress stuff, nothing that's easy to showcase in a nice way. There are some projects I don't even have that to show for it, which is a bit sad.

Anyway, hope you like it. I'm glad to finally have a more up to date thing to point at when I need to explain to people what I do or show someone how well I can do it.

[thepixelcrush]

Detailing in zbrush

[thepixelcrush]

zeedoodles

I wanted to try and see how quickly I could make a little diorama, a self contained- but relatively complete scene.

This was made using only zBrush and Photoshop in about 90 minutes. You can see the twitch live stream archived at the bottom of the page, in two parts.

[thepixelcrush]

Dew

Just before Christmas 2012, I got to work on a project with Aardman Digital's Gavin Strange, creating a 3D version of his Dew character designs. The character was a droplet of water who was the proud owner of an elaborate back story, and a set of game mechanics that involved him changed between materials states from liquid to gaseous to solid.

Gav made these lovely designs and I set about creating their three dimensional counterparts.

It was fun to be tucked away in the little meeting room with pretty much just the two of us talking through ideas and working off our laptops. Seems like forever ago.

[thepixelcrush]

No Leg To Stand On

[thepixelcrush]

SleeveKnight renders

original post: www.thepixelcrush.com/sleeve-knight

[thepixelcrush]

Sleeve Knight final sculpt #wip #zbrush #sculpt #cg #gameart #character #slashedSleeve #

[thepixelcrush]

EGX 2014 and Belonging

Day 1

I made the annual trip up to Earl's Court for the third year running to get a concentrated dose of video games, along with thousands of other people.

On my budget transportation method of choice, outside the Colston Hall at 7:30am, ready to leave I found an empty seat around a table with two other guys who are soon joined by a third. Through a sparse amount of small talk I find out they're headed to the same event, travelling from Cardiff and Bristol. One of them has never been to London, he's heard of a place called Harrods, and remarks how all cities look the same as we pull through the morning traffic. I sit there reading about the horrific abuse the 'gamergate' 'scandal' acted as a catalyst for, while one of them makes a casually distasteful comment about two women with blonde hair driving a mini on the road below us, but I'm too much of coward to call out his bullshit in front of his two friends with an hour of the journey left to go. They don't know I'm wearing this t-shirt underneath my jumper.

Outside the exhibition centre I can see the massive advertising covering its flat face, it takes on a new meaning in a post gamer gate gaming community.

Inside I find Tristan and Nigel, Dan is running the Bertram Fiddle stall upstairs. Its really good to see them all, we lunch together later and catch up.

Virginia

Nigel directs me to a stylish looking game heavily influenced by Thirty Flights of Loving's jump cut narrative structure. Scenes will flip between later or earlier moments with the same character in a way that seems interesting without completely disorienting the player. Its lovingly crafted first person animations give it more of a connection to the world than its cube obsessed precursor. This game wins The Pixel Crush Award for Best Evolution of an Underused Narrative Technique.

Gang Beasts

Off the recommendation of the Crate and Crowbar podcast me and Tristan queue briefly to play Gang Beasts. Its a local same-screen multiplayer game with physics based mechanics. Different coloured blobby characters run around wearing various hats, usually confined to a small elevated arena- including a ferris wheel, window cleaning platforms, some kind of wrestling ring on the side of a sky scraper with an unpleasant looking grinder in the centre. You can jump, you can grab, nothing controls with precision which leads to a gloriously drunken looking slapstick wrestling free-for-all. Like possessed flour stuffed balloons on a stag do. You can cling to most parts of the environment and other characters which can create amusing tug of wars where one player will be trying to manhandle another off a ledge but they're being dragged in another direction by a third player who is also desperately clutching a railing. This game wins The Pixel Crush Award for Best Slapstick Simulator.

Shadow of Mordor

Me and Tristan queue for too long while Nigel sensibly opts to spend his time more wisely. It is a confusing and completely sub optimal way to experience a game with the ambitions that Shadow of Mordor has. You're sat down for 2 minutes while you try and absorb what the buttons do, luckily its borrowed so heavily from other action games that much of it is familiar. The world feels empty. The 'nemesis system' that provides interesting emergent stories to frame every combat encounter is not simple enough to demonstrate in this short window of time. A booth attendant tells me I have 3 minutes left, I start sprinting through the empty world just to try and reach a destination on the map. Now my time is up, Tristan leaves too despite having a couple of minutes left to play. This is a horrible way to experience a game. The only thing you can assess in that time is the most derivative and unremarkable part of the game, surely this can only hurt people's perception of your game? I leave to queue some more, feeling confused and frustrated. This game wins This Pixel Crush Award for Most Poorly Structured Game Demo. Which is shame because I know these interesting systems actually work, its getting a great critical reception.

Alien: Isolation

We queue for a talk about Alien: Isolation, a game that looked promising, and then sounded, like it might deliver, and now appears to be shaping up to be a little bit extraordinary. The 3 terabytes of archived data they managed to wangle from Fox has unearthed original audio recordings that were used as ambient sound in the film's marvellous score, a score that Isolation seems have honoured in the right ways- tying the stems of the mix to the player's stealthiness, and ramping up the intensity of the music based on how aware the alien is of your presence. They also uncovered set photos, character continuity shots and tons of other stuff to base new material on. Its pretty exciting given the aesthetic and design of the alien universe is so exceptional, especially going back to it now with a grainy late 70's veneer over the top. This game wins The Pixel Crush Award for Most Hypiest Hyperbolic Hip Hop Hype Hypocrit.

We tube out of Earl's Court, everyone is done with queues and darkness for the day. Nigel and his walking stick get booted off the stairs by a shop attendant at the Forbidden Planet comic shop on Shaftesbury Avenue, serves him right for trying to rest his sore legs. "My little toe is just one big blister" he points out the next day, not unreasonably. Alan meets us outside and we cross the road to a diner, his local. They serve a fantastic "hard shake" a kind of alcoholic milk shake, I make the most of not driving for once. It is good times.

Tristan kindly grants me use of a mattress for the night and we proceed to not get enough sleep for the next day, talking music and catching up.

Day 2

We queue to get in, I messed up which press passes I thought we had so we both have to queue. Then we check out which queues look tolerable. Occulus Rift, Evolve, and Alien all look unbearably long. I briefly play a space roguelike.

Heat Signature

I've been following Tom Francis' new project with interest because Gunpoint was something I quite enjoyed, and I like the way he talks about games. So I sat down to play it briefly, even though it was brief the game was more fun than I expected based on what I'd seen of it. The click to move stuff worked nicely, quelling my fear of Hotline Miami levels of required skill. There's a "soft fail" system where if you're captured aboard a ship you are ejected from the airlock into space, the game is not over, you just need to try and remotely steer your shuttle to your unconscious body before you bleed out in space, then you're free to continue playing after you've patched yourself up. Its like the buddy system of Far Cry 2 but simpler and more self reliant, it just serves the same purpose of not breaking the flow of a game that is trying to give some proper weight to the consequences of your actions. This game wins The Pixel Crush Award for Safest Bet for Actually Being Good When It Comes Out.

Tristan walked me through the basics of Dungeon Keeper, which is being effectively remade as War for the Overworld and then we head to the Division talk.

The Division

There are many shiny shaders and expensive looking graphical effects. The developer giving the talk uses the word immersion so many times I stop hearing it any more, and it becomes a collection of sounds. He says they're approaching immersion for every aspect of the game, except he never mentions the games mechanics. So no mechanical immersion, you know, the actual interactive part. That's slightly harsh because there do seem be some nice details in the animations and the way the UI is completely embedded in the world. Maps that radiate out across the ground from your character's feet, flashbacks presented as digital glowing point clouds that illustrate a moment in time as you move through the environment. But where's my simulational (probably not a word) immersion? Also why in expensive games is shooting the default mechanic, no one ever questions this. And you know I'm only annoyed about it because the rest of the game looks gorgeous, this just accentuates what a shame it is that we'll just be shooting each other in it. Also the tools for creating the buildings in the engine they use is absolutely fucking nuts, also also- why does it need to be a Tom Clancy game? Its irrelevant. This game wins The Pixel Crush Award for Most Ubisoft Game- Gluttenous and Rote Design With an Interesting Idea Buried Underneath An Enormous Budget.

We all meet up with Loz and go for lunch, a burger has almost become a yearly tradition. Then we go separate ways, Nigel for a powernap, Loz back to the expo, and me and Tristan unwind in Kingston's vegetable market before I catch a train to the coach station. Amid the confusion of the megabus end of the coach station not only does my coach leave without me but I don't even notice it arrive. The I'm called 'Sir' in an increasingly patronising manner, I just want to go home and shower. I'm overly grateful to the lady who sells me a ticket to get out of there on the 20:30 coach.

Listening to an Idle Thumbs podcast on the coach home featuring Anita Sarkesian as a guest I wonder why there isn't a gaming expo more like this, is there one already in the UK? I don't want to queue for an hour to play 10 minutes of Orc Stabber 2014, I want to meet creators and hear them talk about their creations, I want to hear critics critique them. I want them to talk about their creations in a way that is specific to themselves, and there is some of that at EGX and it is the best part. It is however largely outweighed by the rest of it. I can't pretend to enjoy the excited chatter of people leaving a talk, buzzing from the prospect of more homogeneity, and then willing to queue for it.

[thepixelcrush]

Astr0naught

Astr0naught

Front

Side

Back

Turn Around

Close Ups

Fancy Cam

No intro text, no pre-amble, nothing. But if you do want that pre-amble please see a post breaking down the sculpt process here:

The renders where about 6 hours a frame rendered at 4k, I ended having to compromise on the face close up though as vray was really struggling with subsurface skin behind refractive glass at 4k, render times exceeded 24 hours per frame at one point. I did a lot of the base colour texturing in zBrush using the masking tools to pick up creases and details to layer up colour variation and try and bring out the form of things. One of my biggest discoveries on this project was stumbling across a gem of a texture resource called subtlepatterns.com, absolutely tons of  tiling patterns that seem to be intended for web and graphic design, but I found perfect for some of the more synthetic materials I was texturing. I even found great patterns for the leather, mesh, and cloth materials, just enough to layer over a base colour and have a really nice effect.

I call the colour scheme I settled on 'Space Baboon', you only have to look at the show texture above to see why. I also played around with the level of bounce light on the skin, toning it down in order to let more of the scattering to show through. Lots of the hottest highlights were graded, as was the fresnel falloff of the visor to prevent too much reflection from obscuring the face. I made more changes than I expected to in comp, in the end I couldn't get the leather jacket looking how I wanted and changed its colour completely, you can see a work in progress un-composited image below. Finally a healthy dose of depth of field, glint, glow, grain, a subtle sharpen, and chromatic aberration were layered on top.

[thepixelcrush]

Astr0naught

In the slow and ongoing, but usually satisfying, learning of zBrush I decided that working on a character project in my downtime would be a good way to go from one end of the process to the other, learning all the bits in between that I hadn't covered yet. Its taken forever, and even as I render the last few frames for this project I'm looking forward to starting the next and doing it ten times better, just because the first time, you spend hours grappling with the simple stuff often enough. Like why is zBrush's import/export scale so uniquely broken?

Concepting (?)

So I started with a zSphere, drawing out limbs and creating a rough figure. Pretty soon after turning that into a mesh I had this.

Its weird writing this up retrospectively, I mean its always retrospective, but not usually with the whole thing completed and with knowledge of every part of the project and how it turned out. This feels a bit Blue Peter.

I did a fair bit on the helmet and hood areas early on, sculpting in what I thought might look good. Adding little details on the ear casings, almost like large headphone cups, toyed with the idea of adding in a "Britney mic", but it was too early for that kind of detail. Wish I'd come back to that.

I rough in two ventilation tubes using the armature brush in zSketch, its the least thought out breathing apparatus you will ever see, it appears to funnel air from under the armpits into the helmet, or maybe its dead air coming out of the helmet to cool the armpits? But then where's the in take? Or oxygen storage? I'm sure its stashed under the jacket somewhere. I had gathered some reference by this point of what I wanted to aim for for some of the specifics.

Reference

Scuplting

At the start I was sculpting everything using brushes which means everything has a sort of hand made look which is ok, but I could have saved so much time using techniques I picked up later. The rivets lining the helmet for example could have been stamped using insert meshes along a curve, or even alpha stamps with the stroke set to stamp at stepped intervals.

I had the whole body as one continuous dynamesh, guessing that this was a normal workflow and that I could cut up the pieces into separate subtools later. I was sort of right. I found out recently that dynamesh is actually effected by scale, so if you are working at a large scale (whatever that means in zBrush's constantly adapting and context-less workspace) then you can get much higher detail dynameshes at lower resolutions.

One of the tutorials I watched had a huge focus on anatomy, to the point of encouraging the viewer to memorise the names of muscles and bones in order to become familiar with their shapes and relationships to each other, really interesting, and it did help me add little details like the recess on the glove between thumb and hand, suggesting the shape of tendons within- but sculpted using the creases of the glove.

"This is the part where I put in a punchy quote pulled from the article, to exude that glossy magazine style."  -- Olly

I found some problems early on with subtools I had dynameshed had their subdivision levels frozen. Not sure what this means, but when I untoggled this option to unfreeze them I got this fantastic result:

And this:

I never intended to deform any of the of the final meshes so neat topology wasn't a priority, I got this from using the zremesher once I was happy with the basic form I started going in and adding finer details like stitching and damage. As is my usual mistake I took the shortest route possible to the part where I get to focus on the details, I didn't really iterate on the silhouette or shape of the character. If I had done, I might have realised that while the head and helmet proportions were hilarious- it would later mean that the face wasn't very readable from even a small distance. Tiny, tiny face.

Face

 I wanted to make a lady astronaut, so I was doing some research and thought that Valentina Tereshkova had a great face, and an incredible history. I found the image of her from her heyday, that I thought most clearly showed her features. It was the Russian space suits that had this iconic orange colour. I didn't really use the photo for reference for any of the gear on her helmet, but the number of metal bands, tubes and dual microphones(?) is fantastic. Off centre zip too, classy.

The face is definitely the part that went through the most reworking, every time I'd come back to work on the character after a break: the face would look wrong, and I'd go in and make the cheeks fuller, make the mouth bigger, give her a chin. Some of the screenshots below are from later on when I'd started texturing the skin, but you can see the kind of pore and wrinkle detail in the sculpt that I had fun achieving.

Detail

I spent quite a while browsing the fantastic Bad King website looking for insert brushes. That's where I found all the little straps and belt loops to line the hem of the jacket.

I also created a nice stitch stencil that I could stamp along hems, you can't see it so much on the jacket from this angle but its there, hiding.

THERE IT IS! In the top of the glove, that tiny zig zag stitch.

This was my first encounter with the h polish brush, a wonderful thing that acts almost like a sanding tool, shaving off the edges of a flat surface with hard corners. I was struggling to get a separation between materials, from hard plastic to padded polyester. This would have been much easier if I'd known how to use polygroups and panel loops at the time, tools that are perfect for creating nice separate meshes based on sections of mesh.

  I had some classic Boston Dynamics image reference for the leg plate stuff, because I wanted a more exoskeleton-esque robo feel. Or something.

The materials I was trying to recreate was sort of a neoprene wet suit type material for the leggings, with the hard plastic shell flaunted so successfully here by petman -->

I used the insert mesh brushes again to stamp little screw heads across the plate, much like in the reference, to give the impression of floating plates, held together by small rivets. There's also am absolute ton of wiring you can see through one of the translucent casings, which I represented with just a couple of external wires connecting the plates cut out sections.

The valves were fun to do, anything that requires the trim dynamic brush to crush corners is fun to do. There is a radial symmetry feature that means instead of mirroring your brush strokes over one axis, as is fairly common, you can mirror multiple times around an axis. So radial symmetry down the z axis split five ways allows me to sculpt and cut away bolt hole five times evenly spaced with one stroke, which is rather neat. It's also possible to get nice kaleidoscopic patterns this way, as brush strokes begin to overlap each other. I'd thought about the breathing apparatus more at this point and created extra tubes to connect the ones entering the jacket to these front valves, still no sense but at it flows from one place to the next.

Final Touches

Beyond this point it was polish polish polish, and then convincing myself its done. Otherwise it will never see the light of day. As soon as the renders are done I'll write another, shorter post.

"Here is a self aware quote summarising the salient point of the entire article."  -- Olly

[thepixelcrush]

Wax Seal

I had some fun a little while ago in zbrush creating this wax seal concept for a logo.

It didn't get used in the end, but it was a good exercise in using alphas with sculpting to raise the embellishments out of the wax. Another good technique was using the 'mask by cavity' and other masking tools to isolate areas of the sculpt that would attract dust, and texture them appropriately. I then took the base mesh into Maya and created a Vray blend material to get the gold leaf layered over the wax effect, a Vray material over the fast sss 2.

Im still working on my astronaut character and she is really nearly done, I'm rendering at a much higher resolution than I normally would (8k turned out to be untenable so I've settled for 4k). This means renders and file sizes are hard to manage, and I can only really use the render farm when its totally quiet. The projects taken so long that I've already moved out of the honeymoon completion phase and into the "things I could have done better" phase. Though I still can't wait to share her, here's another sneak peak at the look dev process for her face texturing and skin shader.