22 July 2014

Phil Young, Texture Artist

Phil Young is one of our senior Texture Artists here at Double Negative.

On a typical show he will be responsible for texturing the really big assets whilst offering artistic and technical assistance to junior members of staff. With wacom pen in hand and paint bucket tool at the ready he paints the various maps his look_dev artist needs to turn that flat grey export from Maya into something ‘photo-realistic’.  Phil’s recent credits include Fast & Furious 6, Heart of the Sea and Jupiter Ascending.

How did you get into the business?

Originally studying graphic design with the emphasis on multi media, I left university bright eyed and bushy tailed only to be greeted by a world that didn’t seem interested in what I had to offer. Having always been interested in computer graphics, I decided to retrain studying an MSc in Computer Aided Graphical Technological Applications at Teeside University. This helped me get my first job in CG working for a games company as a modeller/texture artist (something I did for six years) before making the jump to television on Gerry Anderson’s New Captain Scarlet. When that finished and with no lottery win on the horizon I tried for a job in film and the rest, as they say, is history having now worked the last seven and a half years as a dedicated texture artist.

Is there anything you wish you had done before you joined the industry which would have better prepared you for your career in VFX for Film?

Having worked in CG for a number of years before starting in visual effects the transition was pretty painless. Having said that I would have liked a better grasp of the terminology rather than staring blankly at my first producer when she asked me if I’d had a chance to look at the plates.

Is there any advice you would give to someone coming into the business?

Learn to walk before you run and don’t be afraid to ask people for help. Starting work at a company like Double Negative can be quite daunting, but do so safe in the knowledge that you’re surrounded by a group of very talented people who will always take the time to offer assistance (capitalise on their experience and it’ll pay dividends in the long run).

What natural skills do you think lend themselves to doing your job?

An eye for detail combined with borderline autism.

Are there any particular training / courses you’d recommend?

Any school with a decent reputation in the industry where you are surrounded by like-minded people with a common goal is a step in the right direction. The course I took was very technical with far too much emphasis on programming and mathematics for someone who only really wanted to build monsters and robots. Having said that, it got me my first job in CG which has culminated in me working on monsters and robots for some of Hollywood’s biggest blockbusters so I think it’s fair to say I’m winning.

The worst and best thing about your job?

The long hours are probably the worst thing but you’re never alone in the trenches and it doesn’t matter how difficult the project, you always look back on them with rose tinted spectacles thinking ‘It wasn’t that bad’. The best thing is definitely the work, it’s always challenging and rarely boring and there’s a great sense of achievement when you see the finished article on the big screen. It can be frustrating making endless tweaks to something but at least everyone is striving for the same thing…to make something amazing, rather than something that’ll do which is such a rare thing to find these days.

YOUR QUESTIONS ANSWERED

From Matthew O’Halloran: Hi Phil, how do you figure out what size textures or amount of textures you need to maintain a high enough texel density across your creations for film? Also do you ever need to use modularity in objects or textures in order to cut down on render times?

Hi Matthew. As a general rule texture resolution is based on how big something appears on screen combined with the complexity of the geometry. A simple background object (that appears on screen in its entirety) might have a 2k (2048×2048) texture, a half screen object a 4k (4096×4096) and a full screen object 8k (8192×8192).

As geometry gets more complicated it will take up more UV space when it is unwrapped so a background digi double will need far more texture space than say a background cardboard box for it to have the same resolution. This means you might go from using one 4k texture for a half screen object to using several.

When working out the amount of tiles you need for an asset first you need to generate some UVs. Once the geometry has been unwrapped take the objects largest UV shell and scale it as large as possible within a tile that will give you the amount of resolution you deem necessary. Then using a checker applied to the geometry scale the remaining UV shells so that the texture resolution is constant for the entire object laying them out across however many UV tiles are needed (we call this process normalising).

As for modularity in textures and objects, overlapping UVs are avoided so although similar geometry might appear multiple times within an object (with the same UVs) it will always have it’s own UV space. If these are laid out in the same locations across multiple tiles it is possible to make use of copy and paste functions in texturing packages such as Mari.

From Arjun Singh: Hi! Do you always create textures in high resolution, irrespective of the fact that they will be used for close-ups or not?

Hi Arjun. As I was saying to Mathew, we try to base texture resolutions on shot requirements. Having said that, packages such as Mari enable you to work at high resolutions just as easily as lower with the ability of exporting high resolution maps at lower resolutions should you so wish. Gone are the days of opening a blank 8k document in Photoshop only to think ‘So this is what Michelangelo felt like when he started painting the Sistine Chapel’!

From Guillermo Andreas Tapia Gonzalez: Hi, I’m an admirer of his work and my most sincere congratulations for that, many films in which visuals have eustedes are my favorite, keep it up.

I have a fan… I have a fan!!!

From Daniel Boucher: Is there a lot of difference between creating maps for Film vs Gaming? Is one easier than the other? How complicated is the task of taking a gray-modeled image from an application like Maya then adding photo-details to it? What application do you use to add the photo-realism? Thanks!

Hi Daniel. Creating the texture maps for games as opposed to film will involve similar processes and packages. It’s how they’re used that’s different. In games they’ll use such things as normal mapping to fake the lighting of bumps and dents whereas we’ll use a displacement map which changes the geometric position of points on a surface.

Of course the technology available to games developers has made huge advances since I was working in the industry but I would imagine they still have far more restrictions placed on them than we do working in film. When I worked in games (back in the golden age of steam) the typical resolution for a single texture map on the PS1 console was 16×16 (256 pixels) with a 256 colour palette. Into this expansive space you were expected to paint anything ranging from a door (complete with handle) to a tileable brick texture… I kid you not!

As for turning that grey Maya export into something photo real, that involves a close collaboration between me and the look dev artist. My job is to supply them with whatever maps they need (usually generated in Mari) to replicate the qualities of the real world object when plugged into the shader. The three most typical maps being RGB (colour detail), BMP (fine surface detail) and SPC (specular reflection). Thankfully there is no specific package that adds the photo realism as yet (that involves considerable effort from both parties)… if there was we’d all be out of a job!

From Max Auer: Do you like hard surface or organic texturing more? Thanks!

Hi Max. I don’t really have a preference between organic and hard surface. For me it’s all about texturing things that have never existed in the real world. By that I mean things that are going to feature in the movie but were never built in a practical sense (perhaps they were only ever an art department concept). When you’re striving for photo realism, photo manipulation is used extensively to create textures (particularly with something like a digi double where photographs are projected directly onto the geometry). I find starting with a blank canvas where I’ll use a combination of hand painting and photo manipulation to create textures far more enjoyable.

From Becca Winter: What do you expect in a showreel from graduates looking to specialise in texturing?

Hi Becca. Obviously, if you are aiming to specialise in texturing that needs to be reflected in your showreel. There’s no point having a beautifully animated character if the textures look like they were painted by a monkey stabbing a keyboard so concentrate on that.

With regards to specific qualities I’d look for in a reel, in my humble opinion (IMHO), the most important thing for a texture artist is scale and attention to detail. If you’re texturing a table, focus on the size of the grain. After all a three foot long table made from the same wood as a six foot long table will have the same surface qualities but over a larger area. Applying the same texture to an object twice the size kills the scale with the end result looking more like something from a dolls house as opposed to the table in your mum’s kitchen… scale is everything!

From Kevin Russell: Any tips for extracting diffuse, specular, bump, etc. from photographs?

Hi Kevin. When you say extracting the maps from photographs are you talking about extracting them prior to applying the images to the geo? It’s important to remove any light information from the images you’re going to use for the RGB but as a general rule I create my secondary maps once the completed RGB has been baked onto the UVs in Mari or Nuke (depending on the complexity of the object).

BMP is typically generated by desaturating the RGB and running a ‘high pass’ filter on it. This gives you an image with a mid point of 50% grey with raised areas slightly lighter and indents slightly darker.

SPC is typically generated by desaturating the RGB and increasing the contrast using a ‘levels adjust’. Note that depending on the colour variations of the original RGB you may have to mask off specific areas within a tile before applying the ‘levels adjust’ to get the desired effect which is a high contrast black and white image across all materials of the object. This allows the look dev artist plenty of room to tweak the values in the shader.

From Veeku Neeku: Do you feel as if Ptex could used more in the industry?

Hi Veeku. I’m no aficionado on Ptex and although they appear to have some interesting features such as no explicit UV assignment they aren’t currently well supported within the software used by visual effects house such as Double Negative.

For someone like myself, where the majority of texturing is still ‘old school’ painted directly onto the UVs, I don’t find the thought of painting incredibly complex objects solely in the orthographic view particularly appealing. I realise you can bake Ptex to UV if the need arises but if this is going to happen (and I see very few situations where that isn’t going to be the case somewhere along the line) you might as well have UVs in place to cover these eventualities.

From Alex Jupp: What was the most complicated/hardest asset you’ve had to texture and how did you break down the task and what was your workflow for building up the texture to finished product?

Do you prefer to use photos, hand paint or a mixture of both to create your textures or is that dependent on the required art style?

For entry level texture artist position what do you think is important to demonstrate in your showreel?

Hi Alex. The most complicated asset I’ve ever worked on would have to be the giant lift that travels through the Earths core in Total Recall. Unfortunately I don’t have time to talk about this (plus it still haunts my dreams) so lets take a quick look at how you’d break down the texturing of a non-specific asset.

It’s a good idea to gather plenty of real world reference of the object you are going to texture. This enables you to study how light plays across the surface which is a great help when it comes to generating your specular and displacement maps. If the object you’re texturing is something that has never existed (say you were given an art department sketch with the brief ‘It’s some kind of metal…just make it look cool’) it’s still a good idea to gather reference of objects with similar material qualities to those you will be painting.

As a general rule I start with the RGB although when texturing something like a plane that is made up of panels I will usually paint the DSP first as the displacement establishes the location for colour variations between panels, panel wobble and where dirt collects.

Splitting everything into separate channels I will work up from the objects surface and paint its constituent parts. With something like a racing car the first channel would contain a base colour. The next channel would contain any colour variation such as go faster stripes with the third channel containing any decals. Working on multiple channels gives you the ability to easily feed them into other maps which means if the go faster stripes have a different specular quality to the base colour, you can easily create a mask from what’s been painted in the RGB rather than having to paint it again (this is also true of the decals that you might put in the BMP to add a slight surface variation). The finished texture can either be baked out from Mari as a flattened image or recombined in Nuke enabling you to tweak the texture considerably.

When it comes to generating dirt and displacement, study the object in a similar fashion as you did for the RGB. With dirt I usually paint an overall surface grime (the kind of dirt that a car would collect from sitting around in the rain) along with a bespoke grime (the kind of dirt a car would collect after it’s been driven through a muddy field at eighty miles per hour). These are typically output as a MSK so the look dev artist can tweak the intensity in the shader although for background objects it’s not unusual to bake the dirt into the RGB. With the displacement think about how the object has been constructed. Is there an overall wobble combined with a more specific structural detail and if so how many maps will you need to paint to convey it?

Even though you’ll have a pretty good idea of the maps you require to get the desired effect working on a really big asset it is a constantly evolving process. Plenty of maps will come about as a result of someone saying ‘Do you know what would look cool’ or ‘We just need something to help break up the specular’.

As for your other questions please see my replies to Max and Becca respectively!

From Govind Krishna: Hi Phil, being an aspiring texture painter I would love to know, When do modellers texture their models “in production” and when is it given to a specialized texture painting?! Thank You.

Hi Govind. With the high volume of very capable generalists here at Dneg it’s not unusual for a modeller to texture an asset they’ve built, it’s usually the larger assets that are textured by dedicated artists.

From Craig Dale: Allo, I think I already know the answer, but has someone else already unwrapped the UVs by the time it gets to you?

Hi Craig. UVs will normally be unwrapped by the modeller and should be good to go. Having said that, I always apply a checkerboard to the model and give it a quick once over before I export it to check the resolution or any distortion issues. If there are any problems with the UVs it’s customary to curse whoever did them before sending the model back… alternatively you can fix them yourself.

From Jessica Lohse: Hi Phil, Do you have to unwrap the UV’s yourself and what are your preferred tools and methods for this? Thanks

Hi Jessica. I think you’ll find I answered most of that question for Craig. If I do need to unwrap an object I generally use UVLayout as it gives great results and it’s very therapeutic watching the mesh unfold (particularly if you’re listening to some mellow music at the time).

From Marque Pierre Sondergaard: How do you keep your eyes fresh, when you’ve been working on the same asset for hours/days? You tend to lose the ability to see things fresh/new/other eyes can see. Any tricks to beat that?

Hi Marque. Don’t be afraid to ask what someone else thinks and if you feel like you’re going round in circles take a break. On John Carter one of my tasks was to work out a method for doing scars that we could propagate across the Tharks. Having spent a couple of weeks going round in circles I thought they looked awful but they were approved by the director first time he saw them… just because you think something looks terrible doesn’t mean it necessarily is!

Do you have a texture painting philosophy (like: edges to creates believability and surfaces to create interest, or disassembling the material into patterns and recreating it with natural patterns of completely different objects – like pitted apple skin for pores in animal skin etc.), and if so could you elaborate on that?

I don’t have a philosophy as such although I’ll always remember the advice the legendary Richard Morris gave me… ‘Just make it look real’ (something that I strive for to this very day). The secret to good texturing is layering… it’s impossible to add all the subtleties you might see on an object in one go. Don’t be afraid to add multiple layers some which might only have an opacity of a few percent (it’s all detail). As for using alternative materials to generate textures I do have a set of dirty plastered walls from which I generate most things.

Could you describe your workflow when starting a new asset? Do you work in details layer for layer, or start in a corner with a defining detail or maybe quickly fill the asset with temporary procedurals or tiled images for an approximation?

I gave a description of typical workflow to Alex if you want to have a read of that. As for procedurals I generally avoid using them unless it’s to reveal a non-procedural texture. It’s not unusual to use tiled images but you need to pay close attention to any visible repeats.

What (texture) artists do you look to for inspiration, and how do you keep improving your game when your skills plateau?

If there’s any one texture artist I look up to it’s Double Negatives very own Alban Orhliac (although this is purely because he’s considerably taller than I am!) As for my skills plateauing, you’ll have to ask my supervisors… I’m pretty sure that happened about three years ago!

Our next interviewee will be announced shortly. Remember to join us on Facebook and Twitter for regular Q&As, the latest Dneg news, recruitment posts and much much more.

See you very soon!