Hi! My name is Darío, I am a Spanish 3D artist living in Madrid.
Since I was a child, I have liked comics, illustration, and art. I was drawing all the time. I was also interested in VFX, animation, and video games. That’s why I found 3D art the best medium to develop my skills.
I have worked as a lighter and generalist, but what I most like to do is characters. In my personal works, I always want to investigate and learn new techniques in all areas, from the modeling process to the render and the final compositing.
Having tested the majority of 3D softwares in the industry, I was watching an increasing number of artists using Blender to do their artworks. Therefore, I thought it was time to test it.
In this article, I will try to explain the process of completing my most recent artwork, “Vintage Gentleman”.
Most of the project was done in Blender 2.90, but Substance Painter, Maya, and Nuke were used as well. It took me about 50 hours to complete it.
Learn more on qarnot.com.
Currently, I am searching for 2D concept art because I want to improve my portfolio with some 3D cartoon characters. When I saw the original 2D art by Max Grecke, I thought it would be perfect to turn into 3D.
Step 1. Sculpting
I started sculpting the head from a cube with multiple subdivisions. Next, within sculpt mode, with the Grab brush selected and the symmetry option enabled, I shaped the cranium and the mandible forms, in order to resemble the chosen concept art.
The sculpting tools in Blender surprised me favorably, though they are in development and they can still improve some things.
Once I had defined the basic shape of the head, I did a remesh to rebuild the geometry, allowing me to work in more detail. I placed the eyeballs in the face as soon as possible, to sculpt the eye sockets and eyelids.
For the moustache, I added a UV Sphere mesh and I modeled it with the Grab brush. Then, I applied Remesher to readjust topology in order to sculpt it easily. The bow tie was built from a hollow cylinder, to which a solidify modifier was applied.
For the suit, I edited a box and subdivided it, saving the sculpting for later. The suit lapels were extracted following procedures similar to those of the eyebrows and hair, using the “Mask Extract” operator.
Step 2. Retopology
In this step, I found some problems related to Blender retopology tools. The Snap To Face tool works fine to stick the vertex to the surface of the sculpted object. The problem appears when I enable the “in Front '' option, which must be enabled in order to gain a better polygon visualization, in the viewport display tab of object properties. This option makes the object always display in front of the others, which is a huge problem when you place the camera behind the object, making the retopology process very discomforting.
The two images above show the issue when you enable the "in front" option in object properties. On the left, the geometry in blue is displayed correctly. On the right, however, the geometry in blue is displayed incorrectly—due to the camera angle, it should be occluded by the high-resolution sculpt.
Another issue is that there is no way to relax the geometry while snapping to the sculpted model. I have tried to use Smooth Brush in Sculpting mode, but “Snap to Face” isn't supported yet. For these reasons, I resolved to use Autodesk Maya for the retopology of the most complex objects.
After retopology was completed, I reimported the retopologized head model into Blender and I applied the Subdivision Surface modifier with 4 or 5 levels of Subdivisions. Next, I added a Shrinkwrap modifier, to transfer all the details of the sculpted model into the new one with good subdivided topology. After that, I converted the object to a mesh in order to collapse the modifier stack. Then I applied “Multires”, pressing the “Unsubdivide” button several times to obtain lower subdivisions levels. In this way, I could sculpt similarly to how I do in ZBrush.
For the purpose of adding more complexity to the hair pieces, I decided to add some curves. Blender has the option to create them with geometry, which makes it possible to see them in render. The option is located inside Object Data Properties, below Bevel tap, by pressing on Round and adding a depth value to decide the diameter. Also, you can control the scale and twist of the curve in each control point, using radius and tilt tools.
Step 3. UVS
Once the modeling had been completed, the next step was UVs. Blender has very basic tools for this task, and they seem a bit old and more rambling than other softwares. First, I selected the edges of desired UV shells in the model inside the 3D viewport, then I pressed the Mark Seams operator, and, afterward, Ctrl + A within face select mode, to select all polygons in the object to display them in the UV editor. The problem comes when you try to use the Unwrap operator inside the UV Editor. It randomly rearranges all UV islands, because Blender has taken into account the selected faces, which are actually the entire model, in the 3D viewport.
I found a workaround to unwrap just one UV island. I select it through UV selection mode, then press shift H to display only the selected UV island, I unwrap it, and then press alt + H to display all the rest of the UV shells again. This way, it is possible to avoid the random arrangement of all elements.
Another issue comes when you select an edge or polygon in the UV Editor; this selection does not display in the 3D Viewport window, due to Blender not having a proper synchronization between those two editors.
Step 4. Posing
Next, I wanted to rotate the head a bit, to avoid the obvious symmetry of the model. For that purpose, I made a very simple rig, then I selected this rig and the head model, and I applied an Armature Deform modifier. I weighed all vertices and I parented all separate objects of the head to the last bone of the rig, with the aim of these objects following the head movement.
Step 5. Shading
After I was finished with the UVs, I started to do some tests with Blender’s render engine: Cycles.
Despite this character being a caricature, I decided to search for high-resolution photos of photographed portraits in order to make the skin.
I brought the FBX model of the head to Substance Painter, the 3D painting software that I have chosen for this project. Foundry’s Mari is a good option, as well.
I painted three maps: base color, roughness, and height (bump).
I started painting the height map; onto the 3D model I projected a map texture that had been scanned from a human head, to get all the pore and wrinkle details. In this way, I could avoid the face’s shader having flat-looking skin.
I got the Base Color texture from a high-resolution portrait photograph, then I cropped a rectangle of 2048 x 2048 pixels and edited the image’s edges to make it tileable. Next, I applied the texture onto the model as a fill layer, with a value scale of 3 inside UV transformation options. Then, I cloned it three times to paint three variations: one reddish for cheeks, nose, forehead, and ears. I created another variation that was bluish for the eye socket and eyelid areas and another one that was pinkish for the lips.
The roughness map is a grayscale texture that is used for highlighting oily areas in the face, in this way the dark tones are used for wet areas, like the lips, eyelids, or ears. For that purpose I used the same tiled source texture which I have used for the Base color, afterwards I applied a color correction to invert it and turn it into grayscale.
After that, once again in Blender, I decided to use the Principled BSDF shader to put in all the textures that I had previously created in Substance Painter. The Base Color was connected to a multiply node, which was merged with the height map, to subtly intensify the pore and wrinkle skin details in the diffuse map. I connected the roughness to a Bright/Contrast node, which allowed me to adjust the brightness of the oily zones in the face. I found out that height and subsurface work better at low values.
Step 6. Lighting
Cycles is set by default with a flat gray color applied to the environment of the scene, which adds a light tone to the renders. I didn’t want any default lights in the scene at the beginning because I wanted to start the lighting step from scratch.
To reset all the lights, I entered into World Properties, inside the Surface tab, and I set the Strength value to 0. In this way, I could start without any type of light in the scene.
Inside the Ray Visibility tab, I disabled Glossy and Transmission in order to avoid reflections in glass surfaces from the environment default color.
I used area lights; the first was the main light that comes from the right side, I called it “Key”. Then I placed two fill lights, both in front of the model; one is located on the top (Upper Fill) and the second is placed on the bottom (Lower Fill). In this way, I avoided extreme dark areas in the render. Finally, I added a strong big light behind and another small light, which is located slightly to the left, in order to highlight the silhouette of the character. I called them “Main Rim” and “Secondary Rim”.
It was useful to review the incidence of each light separately. For that purpose, I made a collection, then I put into it all the lights that I wanted to disable in render, allowing only the single light I wanted to review. Then, in the collection properties, pressing right-click on it in the outliner, I could disable render to hide all the lights in the selected collection.
Step 7. Render and Compositing
Once the lighting was done, the next step was to render. First, I situated the camera according to the original concept art.
For the purpose of getting a clear image without noise, I set a large number of samples (1500) in Cycles’ properties. Resolution output was 2400 x 3400, but I reduced it 50% to test the composition.
At the beginning, Blender’s default composer was great to get an idea of the scene, though I switched to Nuke for the final composite, because it has better tools. I painted the background in Photoshop following the brown tones of the concept art.
The final composition is very simple. It is explained in the image below.
Step 8. Final Touches
Next, it was time to improve the final image a little more. To enhance the cloth aspect, I tested the Blender hair system a bit. I added some fuzz to the suit, the bow tie and the inner shirt. It works well with a very simple setup, but it required some render tests. It wasn’t necessary to use curve guides. In the case of the suit, it has 40,000 emission strands with a length of 0.04. I pressed the interpolated button in the children tab and I set the clump to 0.45.
Each eye is compounded by two spherical objects, one for the inner eye, which has the pupil and white/gray shader, and the other slightly bigger, which has a transparent shader. As usual, cartoon characters must have spot brightness in their eyes, in order to add more expressiveness to them. For that reason, I created a specific shader for the transparent sphere of the outer eyeball part.
It consists of a Mix shader compounded by two shaders; one of them has a complete transmissive surface, the other one is a flat white surface with a bit of emission.
The blend weight of the mix is controlled by a gradient spherical texture connected to a gradient ramp.
This project has been fun and challenging at the same time. I have learned a lot about Blender and I would like to keep improving the quality of my artworks. Blender is a great tool; I expect that developers will continue to improve more features in the coming years. I can’t wait for it.
I would like to thank BlenderNation for giving me the opportunity to write this article.
I hope it helps other people to learn about Blender and the 3D character creation process. If you have any questions about this project, let me know in the comments.
About the Author