Lewis' HLSL Cel-Shader (Gadsby, 2024)
As per the marking rubric for GAR204, primary research is an essential part of the module. Therefore I sought out a friend of a friend, Lewis Gadsby, currently an associate technical artists at Spliced Inc. He graduated last year from Falmouth University and obtained a BA in Game Development: Art, a course that overlaps largely with my own. He agreed to an interview with me, which I recorded on my phone and roughly transcribed below. I tried my best to summarise the key parts of the 35 minute conversation we had, which helped me greatly in furthering my knowledge of the profession of technical art.
Lewis found out about technical art initially through his college course which taught him art, animation, programming and audio simultaneously. He admits this was a stroke of luck as his art teacher happened to discover technical art whilst she was teaching him. He found this topic fascinating, which put him in good stead for his degree at Falmouth where he knew from the start he wanted to do technical art.
Unreal is where Lewis spends "about 95%" of his time at his current position at Spliced Inc. He rarely uses Maya and Photoshop, and is predisposed to Houdini but hasn't tried it out yet. Interestingly, his team use a version control system called Perforce, rather than Git which I currently use on my course. Perforce is better than Git for managing larger infrastructure which is the case for his team. I discuss with him difficulties amongst artists trying to learn Git for the first time, to which he advises that I focus on learning the fundamentals of version control before moving into Git specifics.
I asked Lewis what kind of art he had in his portfolio that led him to getting hired as a tech artist - "I feel like my portfolio was that strong" he told me. His work consisted of 3 main chunks:
He told me he also put some 3D assets in his portfolio just to prove to employers he had experience in Maya.
Out of curiosity, I asked Lewis if he used any HLSL in his portfolio, which he said he had, however he believed employers found it as a demonstration of his ability to learn programming concepts rather than hiring him to create HLSL. "HLSL is hard to debug [...] things like Shadergraph give you a layer of protection" he explained. HLSL is more to help you understand what's going on underneath the bonnet of the latter mentioned.
In this vein, Lewis mentioned the application of Python in programs such as Maya and Houdini, referencing that isolated code execution is easier than actual computer science which requires knowledge of building infrastructures etc. For HLSL and Python, you're writing isolated code which is executed.
My next question for Lewis was concerning his skillset at work - "Which skills do you use in your role the most?" I asked him - straight away he answered "Communication". The ability to interpret and translate between the different layers of the pipeline is essential, as well as knowing how to interpret requests to make life easier for programmers and artists. This is why it's so important to understand what the function of other job positions are in the pipeline so you understand the context in which requests/questions are made.
This relates to what Lewis told me about how he uses Unreal in his current job. He told me it's good to have a brief understanding of C++ so you can comprehend the lower level workings of Unreal when it breaks, to which he added that he doesn't have but he knows who to contact if he runs into those types of issues. "A really important aspect is understanding your relationship with engineers, and you should get as much help from them as you can". This spoke volumes to me, as I've felt like I should be comprehending every complex bit of theory I came across, but at the end of the day that responsibility wouldn't always be weighed on me in industry.
We came back to the topic of Lewis' role at his current job, and he disclosed to me how material editor skills is "The niche [he's] carved out for [himself]". Essentially he maintains a massive master material which has vast amount of child materials, each of which are permutations of the master for ease of use for the artists. The permutations are different versions of the master with chunks of code enabled/disabled through a stack switch since this isn't possible at runtime.
Tooling, the process of creating feature that make life easier for others in the pipeline, is another key part of Lewis' job. Currently he creates his tools with Unreal Blueprints, however other tech artists in his studio create tools in C++ which his company is going to assist him in learning - "I didn't have to know it, but they're going to support me in learning it".
I started enquiring Lewis about which resources have been the most useful for him in learning tech art. He told me that everything (bar what he learnt at college from his teacher) he'd learnt about technical art was from the Internet. He listed off some channels and resources off the top of his head, including Freya Holmer, Shadertoy, Minionsart and Cyanilux. I'd heard of a few of these and was excited to learn more about what they had to offer. Lewis kindly said he would create a resource sheet for me, which has also aided me greatly in research for this module and for supplementing my personal learning of tech art.
The way Lewis sees it, you learn team skills of communication, production and pipeline experience from chaotic collaborative projects, and then the technical fundamentals from resources online. This combination of real-world and online information seems like a powerful mix.
Lewis also mentioned how he's been working on optimisation in his role recently, explaining how his team have been working around the expense of point lights in Unreal due to their omnipresent nature. "People generally think the easier something is to do in the software, the more optimal it is: it's not at all". Their workaround for this is creating a tickbox that toggles cast shadows for said lights, which "shows up this little warning message saying 'Hey, this is more expensive, please don't do this'". He tells me this concept of solving problems before they exist in tech art is important to decrease technical debt.
Next I posed the question - What's your typical workflow for working on shaders? Lewis works principally on 2 different types of shaders: standard pipeline materials and bespoke/VFX shaders. For the prior, his goal is to give artists control, make shaders user friendly and optimised for use in game. For the latter, he says "it's like throwing a bunch of stuff at a wall and just finding stuff that sticks". For him, this manifests as "prototyping a bunch, messing around, deleting everything and starting" which he apparently does around "five times" before he reaches a desired output. He likens this to sketching in traditional art practices.
"Maths fundamentals can give you a huge amount of freedom" Lewis says, as he explains how any maths can be translated into a formula to create a visual output. This is an idea I had admittedly neglected for years through secondary and college, unaware of how intrinsically creative the world of maths can be.
These fundamentals also help build optimisation skills - Lewis states that in tech art, "old tried and true" methods keep returning as "the bar for visual quality gets higher and higher", no matter the hardware capabilities. This results in "archaic rendering techniques and optimisations" that tech artists and graphics engineers carefully craft and often go unnoticed. We then discussed the incredible Quake 3 fast inverse square root optimisation (Nemean, 2020), as well as Acerola's video on the sum of sines for rendering water (Acerola, 2023).
To conclude our interview, I asked Lewis what he wants to look into more in the future. He tells me he's rather content with what he's doing now with his focus on "pipeline, production and working with others" - "All I want to do is help people get a good game out the door". He does note, however, that he'd like to try out sound design at some point, but says that's a complete change in mediums.
I thanked Lewis for his time since the interview left my mind buzzing with ideas about possibilities after university and concepts for this module.
Acerola (2023) Rendering water with sine waves, YouTube. Available at: https://www.youtube.com/watch?v=PH9q0HNBjT4&t=1102s (Accessed: 29 November 2023).
Gadsby, L. (2024) Unity Hlsl cel Shader, Lewis Gadsby. Available at: https://www.lewisgadsby.com/portfolio-collections/portfolio/unity-cel-shader (Accessed: 29 November 2023).
Nemean (2020) Fast inverse square root - a quake III algorithm, YouTube. Available at: https://www.youtube.com/watch?v=p8u_k2LIZyo (Accessed: 29 November 2023).
Commentaires