Making Incredible Hard Surface Models for the VFX Industry
Being prepared is extremely crucial when working in the film industry. We have incredibly tight deadlines trying to be hit by many artists in numerous departments - that all have to work together to deliver the final product. It can be a high-stress environment, so the smoother everything flows between departments, the less headaches and work needs to be redone. Being a modeller, we are at the very start of the pipeline, which means a lot of people will be picking up our work. If we do not follow standard procedures within the studio, we run the risk of work being kicked back to be fixed - which then pushes our deadlines on other tasks, and pushes back all other department deadlines.
I will be using Maya 2018 in these articles, as Maya is the default software used in all VFX houses for modelling. Some do allow the use of Modo and Max, etc., but it is not very common. Ultimately all studios have their own pipelines and propriety tools and software they use. I will just try to simulate the general gist of working in a studio. This is also just my opinion and my workflow; others may disagree with some things I say, but that is all good as long as the work gets done efficiently.
It is very standard these days to receive a scan from on-set, or a concept-model from the art department for our task. I just took one of my old concept models and dynameshed it in ZBrush to pretend to be my scan in the examples.
Setting Up the Maya Scene
One very important thing to remember in 3D is that all the studios work at their own scales in 3D space and we don't work in real world scale, or the scenes would be massive. Usually, studios will work at 1/10th scale. Not all studios work in Meters though; some use feet. Checking your scale is very important, as assets are shared between studios on the same film and the scans come from outside companies. If a scan comes from a scanning company it will usually come with a cube that you can match to a 1/1/1 cube after you have done the math to see if it will match your studios default cube. An incoming scan will never already be at the right scale for you unless another artist in your studio has already set it up.
What I like to do is to import the default human that all studios will have for their scale. It is a good indication of how big your scan or model really is - especially if its vastly off. I also sometimes like to bring in a similar existing asset to double check if I am in the right scale range.
There can be massive issues down the road if something is not made to the right scale. It will most likely get kicked back by the rigger which will throw yours and their schedules out slightly. It is even worse if they do not notice and it gets in to shots with layout and animation, THEN gets kicked back. By this time, even more people's time has been wasted, and all those departments will have to stop working until you can get it sorted and then the rigger fixes their end (It happens).
Almost always your asset will be situated at 0/0/0 - The origin, and facing in the same direction as all previous assets.
The direction faced changes based on the studio - but always stays consistent within the studio on all shows. In my examples, I just chose the default view for Maya with +Z being the front, it is always positive Z or X, not negative. Y is always up and will always be up. If your asset is symmetrical it will be placed directly in the middle of whatever way it is facing. How far forward or back doesn't matter too much as long as it is over the middle in a reasonable way. Most ideal being the direct center is at the origin.
As for height, 0 is always ground level. Make sure whatever you are making will sit on the ground in its default position if it needs to. For my example I have the aircraft slightly off the ground as I plan to make landing gear for it which will touch ground level. The reason we put everything on ground is for when other departments use our assets like Layout or Animation, they can just drop and move the assets where they are needed without having to manually move them down to the ground. This is very important for things that will be used a lot, like background props.
If the asset doesn't need to touch the ground, it doesn't really cause too many problems, it is usually standard practice to still have it sitting on ground level or just above.
Even when working on environments it is quite normal to have whatever part of the environment you are working on at the origin, especially if it is modular. Usually this will be referenced and placed in the correct place by layout or an environment artist at a later date. There are specific times when this is not the case, it depends on the supervisors.
Optimizing Scans for Use
The scans we get can sometimes be crazy heavy - to the point of being unusable in Maya, like 30 million polys that you do not really need. One trick to get around this is to bring your scan in to ZBbrush and to decimate it. Just export your scan from Maya as an obj and bring it in to ZBrush.
I dynameshed my model for my "scan" so please ignore the Dynamesh tab on the side; we do not dynamesh the scan.
If you go to the Zplugin tab and open Decimation Master, you will find what we will be using. What must be done first before decimating is Pre-Processing, I usually just click 'Current', because I have one thing in there. This may take some time depending on how heavy your mesh is. My mesh is incredibly light for the example, but it gets the point across.
After that is done, you can select the quality of the new mesh - you can either type in a percentage or how many K polys it will be. Whichever you alter will alter the other sliders. Once you are happy hit decimate and it should go pretty fast.
Make sure to keep an eye on quality lost in the scan, though. You can knock off a massive percentage of the poly-count but you may have to undo and try a few settings to get the most optimal one. With my example I got it from 1.7 million to 99,000 with very limited quality difference. Usually you can get a 30 million poly scan to 1 million, which makes your life a lot easier in Maya.
Once this is done, just export your mesh as an OBJ and import it back in to your Maya scene. I usually save it as a new OBJ and don't overwrite the original.
Optimizing the Maya Scene
Once we have imported our scan back into Maya and it is the right scale and in the right direction, etc, we are good to go with modelling over it.
I do quite a few other steps, though, to organize my scene and keep the file size down. It does happen where we get emails from IT that we need to clear out some space and this helps manage our usage.
Firstly, I create a layer for control of my Scan. You can do this by selecting the geometry and clicking the circled icon in the image. Doing this will put whatever you have selected in to a new layer. Layers can be found in the "channel box/layer editor" on the right. It is best to name your layer by double-clicking on it, so you know what it is. By having your scan in a layer, you can click the "V" on and off to change the visibility of whats in the layer. I also create a new shader for the scan; usually a colour that is faint, so its very clear when modelling on top what is my scan and what is my geometry.
Once this is done, I select the scan and I go File > Export Selection. What we are doing is saving the scan as its own separate Maya file that we will reference back in to our Maya scene. Make sure it is a Maya file and not an OBJ file. The benefit of this is that every time you save your Maya scene file, it won't actually save the scans geometry as part of it, but will reference in the scan Maya file again when you reopen the file. This keeps your file size much lower, especially if you get in to hundreds of iterations of Maya files for one asset. This does happen and then that is when we get hassled about storage.
What I also like to do is in the layer you can click the empty box until you get the "R". When the R is active, anything in that layer cannot be touched in the viewport - this makes it very handy to work around the scan without accidentally selecting it.
Working with Image Planes and a Scan
Most of the time the scan will get us 90% of the way there - and with looking at on set reference off to the side we can get the job done. There are times when there will be a takeover between plate and CG, where it needs to match very closely, though, for that we line up some camera reference in Maya to push how close we make it. For my example, I didn't spend too much time on it as it is a very tedious process - but has to be done sometimes. It is also much harder to do it on a concept compared to on-set photography because we do not know the image focal length, which usually comes with the on set photography.
To bring in the image planes in to Maya, simply create a new camera in to your scene and go View >Image Plane > Import Image WHILE selecting and looking through the camera you wish. You can look through the camera by clicking on one of your existing Viewport Panels > Perspective > Choose Camera. I always rename the camera to keep it organized, too.
Once you are looking through your camera with the image plane in the background, you just rotate in 3D space trying to line up the geometry with the image. One very important thing to keep in mind while lining up images with a model, is that you need to have the correct focal length on the camera or try to guess what the concepts one is. The focal Length can be found on the right hand side in the Channel Box with the camera selected.
Once you are happy with where your camera is, you have to lock it; if you move it, you have to fix it all again. To lock the camera, select the camera and go to the Channel Box. Shift-click all the options I laid out in the image below, right-click and go Lock Selected. This will stop you from accidentally moving your camera. In the Attributes Editor tab, under Image Plane, you can choose where it is visible. By default, it is all cameras - but I usually switch it to only in selected camera to avoid the image planes appearing in your main view port. I also add the camera to its own Image Plane layer, so I can turn it on and off with ease.
In the bottom image, I am showing that you can select the camera and Isolate Select to quickly go back and forth between just the concept and the geometry on top of the concept.
Working with a Concept and No Base Geometry
This is the least ideal situation, but still happens quite a lot where we receive a concept or on-set photos, but no scans or concept model to go from. The way to tackle this is to import your image plane onto a camera, like above and model very basic shapes. The idea is to stay very loose, as you are going to constantly be adjusting your image plane and your geometry to try and find the right balance between the two.
This is a much more time-intensive process than just having the scan or concept model good to go. A trick I use is to put everything on one of the sides in to a group and Reference it in negative space. This allows you to move and pull one side as it mirrored on the opposite side. it makes it much easier for blocking out the main big shapes and changing your image plane to match. Make sure when you are referencing, you delete the opposite group before adding or removing geometry from the original. Maya is not very good at handing references and it can cause view port glitches this way. If you get a view port glitch, just restart Maya.
This is not really about the Maya, but is something to add to preparation.
It is important to always have your reference visible - whether it be concept, set images or your own gathered reference images.
Don't just try to make things up. It is important to look at real life references for inspiration, especially if your concept is vague. If you are working on a fantasy setting, for example, look up reference of real existing knights and castles; not concept art of wizards and dragons. Say you are damaging Captain America's shield, which is Vibranium technology and doesn't even exist, you'd look for how different metals are damaged in the real world. We draw our reference almost entirely from the real world - don't google 'cool spaceship' and use that.
A massive mistake you shouldn't make: Do not use concepts or images of other intellectual properties as your reference and definitely don't present them with your work. A lot of supervisors may request to see your reference when you present the work for approval.
This is usually what my final ready-to-go scene will look like. I have grouped the things. like the scan, human scale and extra cameras together to keep it away from my geometry. You can get rid of the human if you want, but I usually just leave it around as a good reminder, as you can always just hide them.
When I want to see how my model looks against the concept or reference images, I tear off the window and stick it on my other monitor to compare against as I work in the main viewport. You can do this by going to that Camera's Viewport - then clicking Panel > Tear off Copy, which will detach the window.
After that, save the file, use incremental saving and off you go. Please do not keep saving over the same file. I see it happen so many times where people overwrite the same file, it corrupts and then they lose 2 days of work. You can find it underneath Save As.
Hopefully people learn a thing or two from this. These are just my thoughts and my standard practice at work. I cannot stress how important it is to take the time to make sure things are right. You do not want to be added to that conversation between 10 people in different departments asking why the model is broken or in the wrong place. Any questions just ask below and please spread the word if these sorts of things helped you.