Been studying how to write GLSL shaders, just enough to make sure I understand it enough to avoid any gross oversights in my systems way of dealing with materials and textures and realtime lighting. Starting to get to a reasonable, slightly-past-beginner level where I can put together realtime shaders for tangent space normap mapping, phong shading, drawing textured objects, etc. Im experimenting with all of these on the side and outside of my product. Im also very inspired by Substance Painter which Pedro showed me recently, and it has pointed out a need for me to revisit what I know about PBR shading in the context of realtime approximation. The reason the product is so inspiring to me though is at least 75% from an engineering perspective. Its very inspiring from a user-end perspective, but what catches my eye about Substance Painter is how practical of a product it is to make. It has gone deep on the feature end with things like particle brushes, multi-channel painting, painting in the UV view with texture bleeding, but its architectural scope seems extremely small. A handful of realtime shaders are really doing a lot of the heavy lifting here that, when combined with the texture painting, add up to a wonderful application. Architectural scope is much more important to pay attention to than feature scope, and its something widely-ignored since users generally just see the end features (one of my biggest general complaints about software management). Creating the architecture for an integrated environment, for example, can take more time and thought in the way that it broadens architectural scope than writing several hundred modeling tools which have no impact on the SDK. In terms of architectural scope, my system absolutely dwarfs Substance Painters, and it might not look that way at the moment with the limited features and my basic GUI. And Im not aiming to reduce my architectural scope since my long-term goals need everything I have now, but Substance Painter is inspiring to me in terms of how much impressiveness it can squeeze out of such a narrow scope. Its entirely oriented around painting multiple texture channels for realtime approximations of PBR through GPU shaders. They dont even support anything more than a basic read-only geometry engine (no subdivision surfaces, e.g.) from what I can see. So it shows me a way to go really deep with a very narrow architectural scope while still delivering a very impressive product, and I love that from an economical perspective. I keep this in mind every day now to try to find ways to wrap up a version 1 product without taking forever, since Im likely going to have some skew in architectural focus initially (ex: more focus on modeling and texture painting than character animation).
Posted on: Sun, 21 Dec 2014 06:53:00 +0000
Trending Topics
Recently Viewed Topics
© 2015