Ok, so it's been a while since my last update. Between my clients and moving back to London I'd managed to not do anything on this project since half way through Euro 2012. I'd made a few quickly aborted attempts of Larry's head in modo whilst I tried to get to grips with it but without a good solid block of time to get into it I got nowhere. This time things were different. It was still difficult and frustrating at times but I got there. It's farrrrrrrrrr from perfect but I at least feel like I can use modo now and feel somewhat prepared to produce the big final image. I have to say it was fun to finally be able to sit down and do this.
1) The first step was to make a sphere. I made it with 8x8 segments (generally you want to use as few polygons as you need to keep things quick and manageable) and started making it Larry shaped, keeping symmetry on down the Z axis so that I only have to make adjustments to one side of the model:
Here it is looking vaguely Larry shaped:
It's important to keep the topology of the model as neat as possible, you want smooth straight lines and as-square-as-possible polygons. Polygons that aren't 4 sided throw up all sorts of difficulties as I discovered later. This mini-project is pretty much an exercise in bad topology.
Here I've started to refine the front of the head, making it more dome shaped. I also made the entire head less deep:
^In this screenshot I've selected what's known as an 'edge loop' which is pretty much exactly what it sounds like. It's important to maintain tidy edge loops (many people go as far as color coding them to do this). Here you can see my edge loop is less than ideal. I had to do a lot of tedious work moving individual vertices, edges and polygons one at a time to make my mess look remotetly passable.
Next I start work on Larry's mouth. For this I turn off symmetry because the expression I want isn't symmetrical and if you keep symmetry on for the whole process then the end result can end up looking robotic. I added an edge loop around this area to give me more polygons to work with.
^Here is the basic expression more or less complete. I went for a sort of wry smile thing. I'm not totally happy with it, I think I got caught in two minds and what I was going for isn't clear. The topology of the mouth was a real pain and the cause of about 80% of the problems I encountered.
This next image shows it's not quite as complete as I thought it was, hehe.
I stuck with a fairly head on view when shaping the mouth and as a result I didn't see the polygons get out of whack in the other axis. This shows why it's a good idea to have 4 views visible at one time: e.g. perspective, top, front, right. Until my new PC is finished though I've only got the one monitor so this isn't practical.
As I said, keeping the mouth topology straight proved to be difficult:
Here are two dreaded triangular polygons. These need to be fixed. This is why:
The top image shows the model with subdivisions turned off, the bottom one shows the model with subdivisions turned on. Subdivisions are a way of getting smooth surfaces whilst still keeping the polygon count low resulting in a model that looks nice but remains easy to work with.
Although the triangles don't seem to have a big effect in the first picture, when you turn subdivisions on things get messy. As you can see it doesn't look right, the mouth is thinner where the triangles are. It's good practice to keep subdivisions off when modelling because if a model looks good with them turned off then it will look excellent with them turned on but it's good to turn them on occasionally just to see if anything weird is going on like this. I didn't notice the triangles until I turned subdivisions on. Because of all the problems I had with the mouth I didn't model the whole of the inside of the mouth, for this I can just fill it in during post processing.
When I turned the model around, I was in for a nasty shock:
I still have no idea how this happened. Since the model won't be seen it doesn't really matter but it is a mystery that I don't really want happening again.
Next I start to make eye holes. I deleted 4 polygons (with symmetry back on and subdivisions off), added an edge loop around the hole and then started making the eye shape.
Here the eye socket is practically finished:
Again however, only having one view open cost me:
Btw I think this is pretty much what Larry would look like with glasses. It's not a good look for him.
After some minor correction:
I fill in the holes and his eye sockets are complete.
Next I decide it's time to make the neck. I do this in the stupidest way possible, by just extruding a bunch of faces resulting in the least smooth neck you've ever seen:
I smooth the neck:
Really I should've made the neck much earlier since it's not as much attached to Larry's head as it is part of it. Next time, I guess.
Now I model an eyeball which is a simple sphere squashed a bit in the y axis:
I duplicated it and parented both of them to the main mesh so that they will move where the mesh moves. For this still image it doesn't matter but if I was animating this then it would be essential.
Now I make eyebrows and parent them as well.
Next it's time for what is known as 'UV mapping' which essentially means unwrapping the mesh into a 2D net so that it can be textured. Now since what I want is essentially all flat color this isn't really necessary but it's definitely good practice and gives me more flexibility later down the line so I did it anyway.
This is my first time doing UV mapping in any program, which is probably the least popular element of 3D as it's a tedious but important step with no creative element, needless to say I do it wrong the first time around, this was my first attempt at UV mapping a simple eyeball:
I had tried to unwrap the whole thing in one click, it didn't work, which is what those red bits indicate. The red shows areas where mesh overlaps. Because the entire thing was red I didn't realize I had done anything wrong so I did the same for the other eye and eyebrows and moved on to the body!
Sensibly I split this into a few parts, this is the first:
This is the 2nd:
Note that the nostrils are not highlighted in this group, I mapped them separately with the intention of texturing them black, instead of the green I would use for the rest of the model, they're the elliptical objects in the left of the UV editor window.
Here is where bad topology really
comes into play. Larry's mouth was impossible to map in one or even two goes, I had to do it a few polygons at a time. If this was a high detail area then this would be no good as the texture wouldn't apply evenly and will appear tiled and stitched. One way around this without going back and fixing the topology is to 'stitch' the UVs together one UV at a time. This is tedious, difficult and, for this, unnecessary so I didn't do it. It'sa lot easier in the long run to just get the topology right to begin with, especially if you're animating.
Once the whole net is on the UV grid I had to arrange the pieces to fit in the top right quadrant, the more they tessellate then the more optimized the map and thus the scene will be. I do this and then export the map to a EPS file that I can open in photoshop.
This is the point when I realize what the red areas mean and realize I'd better fix the other maps. I do the eyebrows first, which I have to do it two parts. The first:
I was occasionally referencing the 3D artist tutorial that inspired the look I was going for and that pointed me to a feature in the UV editor I hadn't used yet; show distortion. This shows areas where the texture will either be squashed (red) or stretched (blue). The dream scenario would be to have all the UVs as one piece (so the texture has no seams) with no distortion at all. This is usually not realistic, distortion can be avoided but usually seams are put where they'll be noticed least and care is taken whilst making the texture to try to make them match up.
As you can see my map has a lot of distortion but it doesn't matter. For the next eyebrow I try to fix the distortion and I see some improvement but it's not perfect. As I said though, for this it doesn't matter so I move on.
Now I open the body EPS file in photoshop and fill it with the shade of green I want. The mental image I've had of Larry's color has always been a little more saturated and a little less blue than the shade of color I ended up choosing. It seems that after working on this for 5 straight hours last night I had a crisis of confidence and thought that my original intended shade looked too brash. When I got back into modo and realized I decided that rather than go back and fix it I would just adjust it during post processing. This is bad practice, bad me. Very bad. But it was late and I wanted to get done, I'll explain why it's not totallllly horrendous later.
I also originally forgot to fill in the nostrils with black, making the UV stage utterly pointless, except from a learning POV. I went back and did this messily (it doesn't matter) but I didn't bother applying it! In hindsight this is actually a decision I support but I'll get on to why later.
Next it was time to apply that texture. In the next shot you can see the materials view, I'd already made materials for the eyes and eyebrows. I just applied materials in modo and didn't bother with textures as they simply weren't necessary, making UV mapping them utterly pointless!
I apply the texture and this is what I get:
Next I go onto the 'setup' tab and start setting up lighting and the camera. I don't spend a lot of time on this stage because it's not an area where I have as much new to learn as the others and by this point I have already decided that when I'm done I'm going to try to create the image from scratch again and do it much better, so I can focus on lighting then. For now I move on.
Next is the render tab. I set it so the image is 3500x3000 pixels with a DPI of 300. This is big enough to not look jaggy or pixelated without taking an age to render.
I briefly experimented with some of the HDR (high dynamic range) presets. These are panoramic maps which are used to get accurate reflections. Since I'm going for a cartoony flat style none of these are suitable and I stick with my lighting set up. I set it up to render 4 different passes. Final color output, Alpha output, Ambient Occlusion output and depth output. I'll explain what these are and what they're used for when I get into photoshop.
And so I render it:
The thumbnails at the bottom show all renders that I did. All but the last were quick tests to see how my lighting was looking when I was in the 'setup' tab, which doesn't have the luxury of an automatically updating render preview.
Next I opened up all the passes in photoshop:
Top left is the final color pass which, essentially, is the final image sans post processing.
Top right is the alpha output, which makes it really easy to select Larry in photoshop, I can do it in one click with the magic wand. This helps me when I come to fill in his mouth.
Bottom left is the ambient occlusion pass. Ambient occlusion attempts to approximate the way light radiates in real life, especially off what are normally considered non-reflective surfaces. It will help me get a little more depth to the lighting.
bottom right is the depth pass. This shades the mesh bashed on how far away it is from the camera. This is great for adding things like depth of field in post as it can be used as a mask.
First off I get all the passes together in one document and then use the alpha pass to select the background around Larry and delete it.
Next I made a new layer and use a radial fill with two shades of orange to make a background that will complement Larry's skin tone.
I made another layer and then used filters>render>fibres to make a pattern. I tried to blend this with the orange to give the background a texture but I didn't like the look so I deleted it.
Next I made the color output a little more green and upped the saturation. Then I turned on the ambient occlusion layer, set layer mode to 'multiply' and brought the opacity down to 81% (these settings are somewhat arbitrary, for ambient occlusion passes multiply usually works best but it's a case of trial and error). This gave me some better shadows in the nostrils (which I prefer to just texturing them black) and around the eyes, which previously seemed to be emitting a weird glow. I copy/pasted the original color output and put it a layer above the original. I set layer mode to 'luminosity' which helped to take the edge of the highlights, which became a bit blown out when I fiddled with the color of the original.
I then turned my attention back to the background and this time decided to use a real texture:
Here it is stretched to fit, layer mode set to 'linear light' and opacity brought way down.
Next I decided to use the depth pass to adjust the color balance for no reason other than for practice. I made a new color balance adjustment layer, bringing reds way up and set the depth pass to be a mask for that layer. This made things further away from the camera more red than things close.
This is with the adjustment applied:
This is without:
I didn't really like it applied fully, but I liked the extra definition it brought to the shadows so I lowered its opacity to 60%:
Next, using the alpha pass to select the area I filled in Larry's mouth. I then zoomed in to make some small adjustments and fix a weird area of shadow caused by the ambient occlusion pass:
I then added another texture to the background and fixed some over exposure in the specular highlights of his left eye.
I still felt the background was a little lacking so I added a final texture:
And with that I was done. Here's the final image (not full size, full size is too big to upload to iamgeshack):
I'm not totally happy with the final result. His expression is off and he doesn't look as smooth or nice as he should. This was mostly down to technical issues, it was my first time using modo afterall and it was my first time doing UV mapping, texturing and post processing a 3D image like this at all so I'm not surprised. I feel like I learned a lot yesterday though so I'm pretty happy with how my day went. Now that I feel like I've got to grips with modo I am going to brush up on how to maintain good topology and then have another crack at it. My aims will be:
1) maintain good topology
2) fix his expression
3) light him better, less blown out highlights
4) UV map with no distortion
5) add some cool 3D text saying 'Larry' and make it a poster.
6) do it quicker [optional]
Crucially I want to do this properly, the right way so that I can have confidence going into the final scene, which is a lot more complex. I won't be starting work on that until I've got my new PC up and running which [I]should[I] be Wednesday, but doing this again should keep me occupied until then anyway.
I'll probably start Larry's headshot again tonight. If you've got any questions, comments or criticism then go ahead. Thanks for reading! Hehe, I'm like 40% sure that trying to post all of this will break the thread but here I go.