Technically I didn't learn something new from this video, BUT what makes your videos so interesting for me, is that you are bringing all the technical stuff into a real world context. It's these sentences when you say "in a production like this, you would always want to do it like that...". This is some professional background/meta knowledge one can only have if worked in the industry before. And unfortunately at the same time it's that kind of knowledge that is rarely shared on the internet. So thank you very much for the insights and keep it up!
Hi Bob, great video! I'm architect working on archviz for more than 20 years, and seeing someone that really knows what he is talking about in his expertise field, and sharing really working professional experience is pleasant to see!
Thank you so much Patricio! What's the point of having 25 years of knowledge and keeping it for oneself? I like to share my experiences so that the world can learn from it. :-)
This is a great series of videos! People watching this, don't forget that games are little by little creeping up on what film does, so learning displacement is a must! (Game Terrains are using Subdivisions and Displacement).
One thing worth mentioning is that normal maps contains the offset from normal directly, whereas a bumpmap the engine would have to calculate the offsets from normal based on the height differences it represent. This has several implications: 1) Normal maps are sensitive to UV direction and the map rotation, whereas bump maps are more forgivable for doing random scatter type stuff. 2) Normal maps are insensitive to scaling, a 5deg up angle offset will remain that. Whereas for the bump map you'd had to change the distance accordingly. 3) Normal maps are precalculated normal offsets, making them ideal for gaming assets where speed is key. Whereas bumpmaps are slower to compute but has greater flexibility. 4) Normal maps can represent a big area of offset normal with a single color, whereas a bump map need a continuous grade requiring bigger texture scale and/or bit depth. 5) Normal maps are a nightmare to blend properly, whereas with bumpmaps you can do pretty much whatever you want, if provided the maths are correct just swap it for displacement. Visualizing these, I find it easier to "read the screen values" in RAW/None color mode instead of sRGB/Filmic. Working with bump maps procedurally, I recommend working with them temporarily as displacements, as any errors in whatever curves you calculate will be much easier to spot. When done, reroute to bump map and use the displace node scale height value as the bump node distance (with effect as 1). When dealing with microbumps, I usually do a math node (i.e. 0.1 / 1000) as bump distance. The horrid scales (the default values) Blender use are ridiculous. As for shadows - 30 years or so ago (when recursive ray tracing was a thing), we could at least distort the shadows based on the same math used to distort the normals. While not accurate or competitive to displacement maps, it appears this has lost its appeal. I don't know if it's a limitation of how path tracing works, but I absolutely hate how flat everything looks today. Example: www.realsoft.com/manual/manual/manual/material/vsl/pics/bshadow.gif
Heh, in a way the tables have turned at this point, movies are now the small leagues and games the big leagues, if we are just going by earnings of each industry. Though artists are treated like crap in both, so I guess it really doesn't make much of a difference, other than the sort of workflows that go into each, but I suspect that over time more and more tv/movie CG will likely be done in stuff like Unreal, so the two will likely continue to merge and blur the lines.
@@BlenderBob Sure. Can't wait to see what else you will share. I think you teaching good. Pls keep up the explanations why do you click or using something and why they work or not. These little details are super important. At least I learn a lot out of it.
Hi! Thanks for a great tutorial ang congrats to the 1000+ subscribers. Continue with more videos like those but maybe more in depth you will soon grow 100 times bigger. For this video, I will ask you about Blender as a sculpting tool and the pipeline. There is no vector displacement baking or vector displacement sculpting as I know. This was in Mudbox 2011. What to do with your high res models with undercuts? In Mudbox/Z-Brush to Maya I know. This question continues to EEVEE as a non-displacement low polygon renderer and no Ptex implementation. Why is this important? From Pixar it was important to make their Subdiv, OpenSubdiv and use Ptex to improve and fasten the pipeline but also higher the quality for the animators. If you as a animator can use some thing that look as the high res model in real time the animation accuracy will improve. This is my understanding. Is it missing parts in the Blender pipeline or do I miss some of the functionality in Blender or have I got hold of the wrong end of the stick?
Thanks for the comments Stefan! You can do displacement in Eevee using a displacement modifier instead of a shader but you will need to subdivide a lot as adaptive sampling is a Cycles only feature. I don't really use Eevee as I work in the VFX world an I don't really care about the real time stuff. For vector displacement, since Cycles supports it, you could always do your sculpting in Mudbox. It's like 10$ a month. Or you retopo the model. You can find many tutorials on retopo in Blender. Ptex is cool but hard to implement as it's not supported by all painting software. Substance doesn't. At Method, we talked about it and concluded that pTex was more trouble than anything else (that's what happens when you have unlimited rendering power). ILM and Pixar do use pTex but, they are ILM and Pixar... Blender is perfect for a small to medium size studio but it' not ready for the big one yet, except for the modelling. I hope that answers your questions.
Hey, are you intentionally setting the images that you're gonna bake on to linear? When baking normal/roughness/disp maps I usually set it to non-color, should I change it?
@@BlenderBob Thanks, good to know. I've known for a long time that it should be non-color when importing the textures, but never found much info on whether you should set it pre-bake... By the way, I've been loving all the content, some of the tips you share are pure gold that I never heard from anybody else before, it feels like they're the ultimate industry secrets... So, many thanks! ❤️
You need to convert them first. If you google "convert normal map to height map" or "convert normal map to bump map" you will find many ways to do that. :-)
Absolutely. Just take a look at what Substance Designer can do. It's all procedural. Terrains as often made procedurally. The problem with procedural textures is that if it's not properly done, you will feel that it's procedural. I think it's better to use textures from the real world if possible. And in VFX, you have people specialized in sculpting and THEY LOVE SCULPTING. They want to sculpt EVERYTHING! :-)
J'ai déjà enseigné Maya au Cégep du Vieux et à InterDec. C'était très cool. J'adore enseigner et partager mon savoir mais là je suis CG sup chez Real by Fake et ça me tient occupé à temps plein. Et ça me rendrait fou d'enseigner Blender parce que j'ai tellement personnalisé mes hotkeys que j'aurais de la difficulté à revenir à la version de base.
Ah ah! Not a chance in Eevee. It would all need to fit in the graphic card's memory and I was using 60 gigs of RAM to render. I couldn't even render it n my machine. I had to send everything to the farm to get it rendered.
Go ahead a create a project video for Udemy so we can buy it. Just slow it down, explain the modelling step in more detail like in this video. "I'd buy that for a buck".
