Planning ==== With Artist + Programmer (+ Art Lead + Programming Lead) 1. Look 1. Concept art 2. Example images/videos from other games 3. Photos from the real world 4. Conversation: "Like this, but ..." 2. Budget 1. How much run time (CPU/GPU)? 2. How much data (CPU/GPU)? 3. Preprocessing? Level load? Background ==== How have others done anything similar 1. Papers 1. For rendering: I3D, HPG, JCGT, SIGGRAPH 2. Focus on conferences with real-time focus 3. Gems, ShaderX, etc. books 2. Presentations 1. GDC (some free in Vault, check for slides elsewhere) 2. SIGGRAPH courses 3. NVIDIA GTC 3. Blogs 1. Google 2. Game Developer (but doesn't always get technical enough) 3. Game developer's twitter 4. Graphics weekly 4. Is there anything that looks right, but is too slow 1. Film Production 2. Physics Simulation Prototype ==== 1. *Fast* iteration, fake assets 1. Small test map or sandbox app 1. Many developers build a sandbox app outside game engine 2. I used to use one that auto-reloaded any shader file or image file change, w/ sliders for shader parameters. It could load static models but not animation 3. For shader only, I've even used shadertoy 2. OK to test on plane, cube, sphere 3. Test optimization (if necessary) 1. Approximate, bake into texture, leave stuff out 2. Will do final optimization later, but need to have some confidence the final version will be able to achieve the look you want (especially if approximating or dropping some parts) 2. Mock up effect, share with artists 1. Feedback: is this going in the right direction 2. It won't look final, try to look past that. Implement in Game ==== Programming step 1. Leave control to artist 1. Textures with as plausible as possible default 2. Parameters 1. With plausible defaults 2. And limits on values 2. Reparameterize texture parameters * Should be 0-1 3. Make a preprocessing stage on import or load if necessary