Hello everyone:
I recently sent in a suggestion for a new course and got a great response. It was also suggested I throw it out to the community to get some additional feedback. Here was the bulk of my original email:
"I'm a hard surface modeler at heart and always will be. However, I've never been good at putting my models into scenes to give them context.
This is particularly true when it comes to using characters. CG Cookie offers amazing tutorials on detailed character generation but what if the character is not the star of the show? What if it's just meant to bring life to a static scene?
For example, I can model a steampunk vehicle, but I'd love to have a scene showcasing that vehicle with perhaps a mechanic working on it or using it. I can generate AI characters easily enough and Photoshop them into a rendered Blender image but it's just not the same.
I'd rather do the whole thing in Blender but need a dialed-down course on how to generate a character just for this kind of use.
I guess what I'm asking for is a realistic (not stylized) character creator course for intermediate hobbyists like me who have little to no sculpting skills. The focus of this "Character Creation for Hard Surface Modelers" course would be to generate characters to use in scenes from a distance. No close-up rendering.
This is probably vague but wanted to throw this out to you anyway, for what it's worth. Thanks again for everything you do."
This was the response:
"Hi Jonas - I think I'm tracking with you, and it's a great suggestion. Filling out our renders with everything that's not the focus is important but rarely taught.
As I envision a course covering this subject matter, I see two ends of the spectrum: A) just build everything the hard way and B) cheat everything. The former is kinda obvious and wouldn't warrant a course imo but it would take forever. I'm guessing the latter is really what you mean: How can we do this quicker than building it all the hard way?
There's only one reliable quick solution in my experience:
Use images for still renders. With a good choice of environment and some basic compositing skills, this is the best choice for speed and quality. Images can still be used for slight animation but it needs to be subtle. I demonstrate this in the presentation chapter of the HUMAN course, sourcing public domain photography sources.
AI is a polarizing right now but you certainly can generate imagery to serve the same purpose. It's probably the best option if you want characters in your scene.
Aside from imagery, the task of filling out a scene becomes wide-ranging. Very difficult to describe which is why it's a good course suggestion!"
After giving it some further thought, my desire to create a basic scene for the sole purpose of showcasing one specific model goes beyond character creation. Even if it wasn't possible to get a good quality character by cutting corners, I think it would be reasonable to model a small scene with or without characters that could be re-used to showcase as many models as I want. This is difficult for me to describe in words so I apologize if I'm not being clear. As an example, if I wanted to model an antique car, I could also model a basic garage to put it in - almost like a diorama. Because the focus is the car and not the garage, I would use quick-and-fast techniques like image planes, projection mapping, etc just to get the job done for a well-rendered still image.
Thanks for that. I agree addons can save a lot of time and I may end up using some as part of my solution. However, I’d rather come up with an idea that isn’t so heavily dependent on them. Addons come and go.
One idea I may try later today is to bring in an image of a character with a transparent background, then use a base mesh posed in the same way directly behind it but only have the shadows from the mesh in the render. So, in theory, I’d have the photographic detail from the flat image plane, but 3-dimensional shading. It wouldn’t be perfect, but that’s the whole point.
Oh yes, I can feel your pain. One of the things that is always tough is to model something cool, only to find yourself thinking, where am I going to put it? And then it strikes you, I have to model a whole environment now? Oh my god, that is too much work. I remember Kent doing a turntable platform in Spice Vendor Collab, always found that a nice solution.
https://cgcookie.com/live-streams/collab2021-week-5-closing?playlist=7byq7x
I think you captured what I was thinking. I foresee a workflow that relies heavily on techniques like projection mapping which gives lots of photographic realism with minimal effort. Chris Bailey’s Bug Bots course is great for that. In the end, I can see an instructional course combining half a dozen or so down-and dirty visual “cheats” intended primarily for still images.
This is really close. This link from Purple Park Studios is a couple years old but the principles apply. A number of issues need to be addressed. First, if the feet are going to be showing, the shadows have to account for the difference in height. Second, an image plane has a fixed lighting direction so the lighting in the scene can’t conflict with that or it will look odd. Third, shadows in real life have perspective and this method by itself doesn’t account for that, and finally, any shadows cast by other objects in the scene onto the image plane will be completely flat and destroy the effect. Other than that, it’s perfect, lol. I know it sounds like I’m talking my way out of this, but I’m not. I still think it can work.
Putting characters aside for a moment, I think I want to focus on the structure first and see how that goes. Here are some images I generated for reference. I wouldn’t model directly from them but it gives me a direction in terms of overall look. I particularly like the one that has an open area to the sky. I’m also thinking that maybe some of these elements could be modular. The goal is maximum detail with minimal modeling. As I create my first attempt, I may open up a project in the gallery as a work in progress.
Sounds like what we did on the robot Collab. Kent made the raw livestreams available if you want to take a look:
Thanks, Omar. I'll definitely take a look. I was on board for that collab but had to drop out early due to serious family obligations. I'll take a look at them before I get started.
