Modelling a Tyrannosaurus Rex: Zbrush associated to Maya.

With this making of, I’ll present you all the steps of my work, which enabled me to get the sculpted final model of the T-Rex.
I will also try to group all the conclusions from my researches and tests on the conception of a high poly model thanks to the union of 2 tools: Maya and Zbrush.
Indeed, I had to learn the modelling techniques of Zbruh and the way of exploiting them on Maya. Since the “David” and some other projects, I’ve improved my way of getting onto 3D modelling; I found it frustrating and little intuitive, moving points during hours, trying to get an acceptable shape. I’ve always preferred art to technique, but today, the two “facets” tend to mix together. I now work with a more creative method: artistic and productive, linking simply art and technique, and in the same time dissociating them: the process combines 2 softs: the first one is used for render and animation (Maya), the second one for modelling (Zbrush). The work is organised in different phases, either more artistic, or more technical. This enables me to have an accurate view during the advancement of the model, in the meantime regardless of both “facets”.




The initial step is to get familiar with the shape of the T-Rex, its morphology, and try to identify the distinctiveness that you want to stress on the model so as to give it a minimum of reality, but also so as to personalize it. During these researches, I select about 30 images of a T-Rex, and, although my model is the one of Jurassic Park for a while, it enables me to improve the appearance of the T-Rex that I want to model.
Since I don’t think that I’ll be able to get an extreme level of details, I’ll bring out what I like in this animal.
Another point: since nobody ever saw a tyrannosaurus, I know it only thanks to human imagination. That’s why I want to have images of its skeleton, the only real and irrefutable vestige... This point is really important to have a quite reliable basis of the proportions or of the number of teeth for instance.


Then, I draw so as to set straight the features that I like and so as to « learn » my model, since my visual memory is not enough.
This step is more or less long, it depends on the time you have. I am not hurried and I like this dinosaur to much to go quickly (of course, you don’t have the same state of mind when you’re at the end of a production ;o) ) so I enjoy and also feel really impressed, nearly worried, to have to duplicate what I’ve always admired.

--- In Maya ---

Initially, I always begin with a low poly draft that I create in Maya. I model very simply the general shapes of the model and a posture. I don’t stay too long in Maya, since all the work for the volume and structures will be done in Zbrush, with a simple graphic pen.
Actually, I prefer sculpting a very low poly in Maya, but this first step could also be done with the zspheres of Zbrush.
This very low poly doesn’t to need to have absolutely square faces. It must have minimum details (I process to a re-meshing later). It will be then easier to place the specific elements of the model, such as eyes, etc, with Zbrush.
Once finished, it Is better to be sure that the model is perfectly symmetric before exporting it.


--- In Zbrush ---

This step consists in modelling without to many details the low poly created earlier in Maya.
I import the object to sculpt in Zbrush (Tool – Import); I slide the mouse so that the object appears in the working space and press « Shift » before releasing the left button of the mouse (so as to axe the model).
I immediately activate the Edit mode (T) and activate the symmetry (Transform), and then I begin to retouch the T-Rex with the Draw (Q) and Move (W) modes.
You have to know that the model I want to have in this step is not the definitive one, but just a 3D draft, representing the mains specificities, proportion and dynamics of the final T-Rex. Actually, what I want in this step is to get quickly and intuitively the volume of the animal. This work will be useful in the next step, which is more technical.
The “Move” mode is very useful to modify the shape of the model, its thickness, but not to elongate or shorten it. So as to elongate or shorten it, I use the lattices of Maya. I go from Maya to Zbrush and Zbrush to Maya many times during the production.
The “Draw” mode is essential so as to go into detail for the shape. I prefer the “Inflat” option to the “Standard” one, even though the difference is minor (besides, it took me a long time before seeing the difference!), but I have the impression that the pixols “more properly” react when I use “Inflat” and they react in relation to the normals of the model, and not to the camera… Anyway, the “Smooth” option is crucial: I define the parameters of intensity, and then I constantly use it by pressing on Shift.
I subdivide the model when I feel I can’t go further with these pixols.
During this step, I work maximum on the level 3 or 4.
I then export the model with the highest subdivision level (3 or 4).

--- In Maya ---


After having created the volume of the T-Rex in Zbrush, I import it in Maya .Thanks to the “Make life” option (Modify – Make life); I start the definitive modelling of the low poly T-Rex. Definitive meaning that all the work of this step consists in re-meshing the model so that the mesh is as efficient as possible, so as to be a support of the future displacement map and to be skinned and animated correctly.
The advantage of this production method becomes meaningful in this step: now I don’t care of the volume since the polygons are snapped on the detailed model in Zbrush. I focus on the coherence and efficiency of the meshing.
Since I now have a real model, I can model face by face, while drawing on my model the dynamics and articulation which will follow the polygons. This is the principle and advantage of the “modelling face by face”, successive extrusions of 4-sides faces according to a simple plan: the dynamics of the model.


This phase 3 is although a challenge, which remains of course really simplified by the modelling technique used: having only 4-sides polygons… squares everywhere. Why? Well, without being an obligation, I thing there are 3 reasons:
. Once the re-meshing is finished, you can export a clean model that Zbrush won’t re-interpret by changing the topology of the meshing.
. You can generate a cleaner displacement map.
. You can optimize the render time under Mental Ray.
Of course, wanting to have only quadrilaterals is interesting when you know that you will use a displacement map to create details. Having 4-sides faces is ideal for animation too. At the end of the re-meshing, you enter in a real logical game so as to place the last 4-sides faces… but don’t worry, there is always a solution!
(I export my meshing in Zbrush so as to check that it has no more that 4-sides faces. If it is the case, a mistake message appears.)

Once the meshing is finished, I watch my model and improve the vertexes so that each line is supple and I cut where it is needed (always keep in mind you want to detail in Zbrush and animate your model). Indeed, when I subdivide the model in Zbrush, I’ll need that each part of the body has about the same number of faces so as to be able to create details uniformly. However, the head must be more subdivided. Finally, each articulation must have enough faces so that it can move correctly.
I then improve the proportions with some lattices. I delete half of the model and recreate the symmetry (Polygon – Mirror geometry) after having aligned the vertexes (Discret Scale) which must join. I apply a “Merge vertices” on all the vertexes (with a minimum distance of 0.0007) so as to be sure that it acts only on the snapped but not joined vertexes.
Last step before Zbrush, I create polygroups which will be useful for isolating easily each part to detail in Zbrush (Create – Sets – Quick select set), especially for the inside of the mouth and the head which are difficultly dissociable without this famous « Quick select set ».
For the moment, I don’t model eyes, teeth or jaw; they will be separate elements which will be added latter.
Before exporting, note that the number of polygons of the final low poly model will be definitive. Once you start working the sculpture on Zbrush, you can come back in Maya, whenever you want, to move the vertexes, deform the model, redo the UVs BUT you can never add or delete faces.
I export my clean low poly model ready to be detailed in Zbrush.

--- In Zbrush ---

      Welcome in the most abstract, the most artistic part of the project, where observation, training and perseverance are the most important!
I will describe my way of proceeding: from the import of the object to the generation of the final displacement map.
I import the T-Rex in Zbrush just like in the second phase. Once the Edition mode is activated (T), I care about the UV of my model. I will use the automatic UV unfold for 2 reasons :
. First, this gives me the assurance of having no deformation of the texture and especially of the displacement map.
. Secondly, it works! :o) And it match with my way of working: we don’t need a manual UV unfold, since all the produced textures for this project are generated in Zbrush, and not in a 2D software such as Photoshop.
So: Tool – Texture – GUVTiles.
The UVs can be unfold at the beginning or the end of the modelling work but they must be unfold before the use of « Projection Master » (I have not Zbrush 2.5 but the note must have no sense now since “Projection Master” doesn’t exist as plug in).
Before beginning to sculpt, I activate the “Multi Draw” (Preference – Performance – MultiDraw). Since my processor is a dual core, the benefit is real.

I subdivide my model at level 2 (so as to be conscious of the next threshold), BUT I start working at level 1 because I thing that the 2 first levels are the basis of the structure, proportions and of the future features I want to stress.
I take time, principally in the « Move » mode. The « Frame » option activates what enables me to visualize the points and dynamics of the model. With the few pixols I have at the moment, I « prepare » the areas which will enable me to emphasize the creases and muscles.

At level 2, I keep on developing the muscles. The thighs and calves and the powerful tail of Mister T-Rex. I have fun in emphasizing the bony parts, especially the one of the pelvis and of the head, making standing out the archways and their wonderful protuberance, stressing the top of the head, so typical of the T-Rex.
Of course, I always look at my graphical resources, so to place each element correctly. I print the one I prefer so as to see them every time, to be always impregnated with my model… particularly when you live in a so little flat!

   Until the level 4, I may say that I only care about the anatomy of the T-Rex by modelling the muscles, the bones and the fat that I want to let appear, meanwhile I keep in mind that skin covers the whole.

 The fat… wonderful, when it is combined to muscles or to the bones! That’s why, from the beginning, I had the idea of giving a flabby and vulnerable aspect to each part situated under the animal: from the throat to the tail. I wanted that this “tyrannical” dinosaur doesn’t look like a monster but like a living animal.
The upper side of the T-Rex will be rigid with bony excrescence. I want to stress this property on the upper side of the pelvis, thinking to whale which seem to have real rock on their body. By splitting early those 2 areas, I know that I will have to work differently on the upper and the lower parts of the body.

The 5 and 6 levels are dedicated to the soft and rough aspect of the body. The legs are rigid and protected by many bony plates. The claws will be worn. After having considered it, I do all the details manually with the « Draw » mode, with « Inflat » then « Standard » for the most little and irregular details. I use “Projection Master” only for the legs with the very useful tool “DecorBrush” (in “Zsub”) so as to create scares.

At those subdivisions levels, it is necessary to isolate the parts to sculpt (Shift+Ctrl+left button of the mouse), because, particularly at the level 6, the number of pixols is difficultly manageable (5 millions) but also… so as not to panic on account of the work to do. It is easier to concentrate on little areas and to organize your work time in “areas” to detail. Generally, I know that I will come back on all the parts at each level of subdivision, until the level 5, so you don’t need to spend too much time on each part.


Once the level 5 is finished, the volume and the proportions of the T-Rex have changed a lot. I then come back to Maya, taking care of importing the new model (level 1 of subdivision) by nicking “False” (Import Options - Create Multiple objects) so as to be able to find my work in Zbrush again. With some lattices, I make increase or decrease… until I am totally satisfied with the proportions.
Back in Zbrush, I import the model that I have retouched in Maya, over the last one and, the changes are taking into account! By increasing the subdivisions levels, I get exactly the sculpture details.

    At level 6, this is the final step: I place all the details which were difficult to put before (vein, pimple, etc.), and I also soften or exaggerate some reliefs. Then, I check that all the detailed areas match together before stopping the sculpting work.
Here comes the moment of generating the displacement map!
I come back to the level 1 of subdivisions.
After innumerable tests, I’ve adopted some, according to me, important set up:


. First, since my displacement map will be used with a model in polygons in Maya, it is important to activate the « Cage » mode (Tool - Geometry) (The meshing explodes, this is normal but it will be clean in the render with Mental Ray). Thus, you can get a model in Zbrush more similar (but never identical) to the Zbrush version when you render in Mental ray. Indeed, without the “Cage” mode, your model will get deform and let appear the details of the displacement map.

. Then, I define the parameters of the displacement map (Tool - Displacement) : DPsubPix at 2, DPres at 4096 (resolution of the displacement map, at this level, the map will contain all the details of the sculpture), SmoothUV activated (only IN 3Cage3 mode, otherwise the displacement map will be incoherent), then « Mode » activated.

.  Finally, in the options of export of the object (I do it now but I’m not sure that it will influence the displacement map), I let off « Grp » and I nick « Mrg ».
That’s alright: ‘'Create Dispmap'’.
A few minutes latter, the displacement map appears in the windows of the alphas. I select it and I “FlipV” so as to orient it so that Maya reads it correctly.
I export the dispmap in TIFF, and then I export the T-Rex subdivision 1 in « Cage » mode and open Photoshop.




Once the sculpture is finished, the challenge is to recreate the details that the sculpture has, on a low subdivided object, because the goal is to skin and animate the object!

--- In Photoshop ---
This phase begins with Photoshop (CS) a few minutes, this is the time to discover the displacement map generated in Zbrush. Before opening the dispmap in TIFF, I define the parameters of Photoshop: the soft must receive the dispmap so that it is as efficient as possible once in Maya and so that there is “distension” phenomenon during the render.

    In: ‘’Edit - Color settings’’, I let off the colour management by putting « Off » in « RGB », « CMYK » and « Gray ». Then, in the same window, in « Working space » at « Gray », I set « Dot Gain 30% », i twill darken the map of 30%.


Then, I open the dispmap in TIFF and convert it in RGB 16 Bits (Image – Mode) so that it can do its job in Mental Ray. I increase the luminosity of 3 points (Image - Adjustements - Brightness/Contrast).


 Generally, the map has some defaults, kind of black and white spots. They will influence the render by giving some errors, so I use Blur Tool to soften them.
Then, I back up the dispmap, always in TIFF.
--- In Maya ---
Maya, here we are!
I import the T-Rex exported from Zbrush, always with the “False” option nicked.
You have to know that the size of the model has no importance; it only acts directly on the intensity of the displacement. It means that the bigger the T-Rex is, the more you have to increase the value of intensity of the displacement so to obtain the wished details.
For my render, I’ve used a shader of skin but the principle to do a displacement is the same than with a « Blin » or a « Phong ».
Construction of a shader aiming at creating a displacement:
Just 3 elements of the Hypershade : a « Blinn » (or a « Phong »), a « displacement shader » and a « File ».
This is the scheme of the connexions:

Then, I double click on « File » and I import the displacement map. In “Color balance” with the “Alpha Gain”, I set the intensity of the displacement and so of the relief generated at the render. The higher the alpha gain is, the stronger will be the relief and details.

However, you still have to set an expression in « Alpha Offset ». The aim is to automate the fact that the Alpha Offset have to be equal to alphaGain / (-)2.
This is the expression to put in “Alpha offset”: =-file1.alphaGain/2 (put a capital letter G and the right file number. Here: file1).
The shader is ready. I associate it to the T-Rex. Then, I select the T-Rex and open the “Editor attribute”, in the options, there is a panel « Displacement map », and let off « Feature displacement ».
Well, so much technique… just one step before beginning the render tests.
We know have to associate a “displace Approx” and a “subdiv Approx” to the meshing of the T-Rex. It will make the appearance of the displacement map more efficient when rendering (with Mental Ray). They are in: Window - Rendering Editors - Mental Ray - Approximation Editor. By selecting the T-Rex then by creating a displace and a subdiv approx, they are assigned to the meshing. We just have to edit them now. Here are the parameters:


So as to find the right value of the AlphaGain and consequently of the displace, I do many renders on the head for instance, by deleting the rest of the body to decrease the render time. I then compare the different images and chose the one I prefer by reminding of the AlphaGain value associated to it.
Then, I model the tongue of the T-Rex. I start with a low poly version (a few square faces) that I sculpt in Zbrush so as to generate a displacement map, useful to get the details.

Exactly the same principle as for the body (except that I don’t go through the very high poly phase, since the tongue is easier to model properly). Then in Maya, like in the phase 5, I create a new shader for the tongue with the displacement map associated to it, without forgetting to associate the earlier set up displace and subdiv approx to the meshing.

Finally, I model the teeth. Since I want accurate details, I will proceed the same way. The sculpting work will be done on just one tooth. So I will have only one displacement map that I will associate to all the duplicated teeth. The placement of each of them is important so as to give a style to the jaw. The reference images are welcome!


Also, since the models (body, teeth, tongue) that I place in my Maya scene are imported in « Cage » mode, it is difficult for me to place precisely the one in relation to the others. That’s why I place the teeth and eyes on a model of the body at the level 3 of subdivision (without activating the « Cage » mode), its smooth aspect enables me to be more accurate.

So, this long Making of is finished, I hope I’ve been clear enough in my way of explaining my work. Should you have any question, do not hesitate to write me an email at the following address:

Thanks for your attention.

I'll describe the render process another time (mental ray, occlusion etc). Anyway, if you want more informations on the shader, do not hesitate to read my making of about this.