Enrico Caruso

This topic is about my next project. It is about a singer I love.

From time to time, I also will give some feedback or share some thoughts.

Happy blending.


Lattices not for deformation, but they are useful to quickly measure and sketch proportions.

Voxels are used for sculpting. I need to split the body into different segments to prevent lags later. (Adaptive Remesh sadly gives poor results, not a good idea to rely on it.)

Body segments.

Using silhouettes. There is the grap brush, the snake hook or the elastic brush. But since we do have “projected” falloff, many more brushes become useful for a silhouette based workflow.

I used flat shading to get rid of distracting shadows and keep the focus on the silhouette.

Happy Blending.

1 Like

After working with contours I go on with regular shaded sculpting.

1 Like

Current state … I did not work on hands and feet yet.

1 Like

Some updates

1 Like

Voxel sculpting legs …

Some thoughts …

When used with projected falloff, “Draw Sharp” and “Scrape”, both could be very good contour brushes.

Using “Auto Smooth” together with “Normal Radius” is useful to smoothen the surface, and it is useful to avoid bumpy sculpts.

A high Normal Radius helps to average out bumps.

The Flatten brush works very good with Auto Smooth and Normal Radius. Slightly increasing “Plane Offset” (Auto Smooth active) gives us a Bubble Smooth brush.

I tried the “Clay Thumb” brush several times. Still, I do not get good results with it. It would be interesting to see someone working with it.

In my voxel sculpting workflow, the adaptive voxel remesher is the weakest part. It does not give reliable results, creates tons of artifacts on low-slope areas and erases details. Because of this issues –– sadly –– I have to avoid the adaptive voxel remesher completely.
I mostly keep everything in separate meshes to compensate for this problem. But, since we could quickly switch between meshes from Sculpting mode, this workflow does work well.

1 Like

Years ago, a classical sculptor told me an idea which helped me a lot to sculpt hands.

It was that simple:
“Sculpt a thumb, as good as you could”. After mastering this exercise, my next exercise was “sculpt a finger”.

After I sculpted several fingers and thumbs my hand sculpting improved.

This time I merged most meshes (except fingers) to a large mesh, risking laggy workflow. But, to my happy surprise, most sculpting brushes work fluid. I do not have much experience with face sets, so I tried them now. Arms, legs, head, abdomen, all have their own face set.

Hiding face sets turned out to be very useful.

1 Like

Hand got updated.

After I merged fingers to the body, I had to increase voxel remesh density, to prevent fingers from merging together after a remesh operation.
The model now has 20M faces. Sculpting still fluid here, but some operations (like saving, remesh or other operations) take a considerable amount of time.

To achieve details nescessary for hands, eye lids, finger nails or other details I also have to create excessive geometry for areas where detail is not needed. For this reason, I believe, a working adaptive voxel remesher could be one of the more important missing parts for a voxel based workflow for realistic sculptings. It helps to keep details where they belong, and avoid unnecessary details where they are not needed.

I also need to say, the Inflate Brush (Projection Falloff) is a very good contour brush.
Face Sets pay out. I can hide large areas in no time. The concept of Face Sets also encourage me to dare to sculpt the mouth cavity. Because I can use Face Sets to quickly hide everything not related to mouth cavity.

During my last realistic sculpting projects I can confirm, that Pablo Dobarro’s work on Blender Sculpting improved Blenders realistic sculpting capabilities.

Happy blending.

1 Like



1 Like

Retopology of hands.

Some thoughts …
I excessively used “Poly Build” tool for retopology. On long shapes, like arms or fingers, I extruded long cylindrical shapes, and cut segments with the knife tool.
Shrinkwrap used together with Snapping (Face).

There seems to be a quirk with the Poly Build tool. It tries to grab geometry which are backfacing. It happened several times that I accidently extruded a back-facing edge or moved a backfacing vertex (and then pulled to front by snapping). On some moments it was difficult to select front facing edges, with PolyBuild mostly selecting backfacing geomentry. These moments I end up rotating and panning the view, to get enough tiny area on the screen to extrude the edge I want to extrude.
“Shading // Backface Culling” does not change that behaviour.

Some thoughts on Poly Build:
It could be improved by ignoring backfacing geometry. Or an option which disables backfacing. So, a user only acts on frontfacing geometry.

Working on retopology of the fingers, I had a retopomesh for each finger. And each finger was an object. It does enable mirroring for fingers. And I can align the view to each finger easily.
To align a retopology-finger to the sculpt-finger, I still rely on “Copy All to Selected” for Location and Rotation.

Another idea:
If I do Snap “Cursor to Selected”, target object rotation is ignored…
And “Snap Selection to Cursor” does ignore Cursor rotations.
It would be nice if snapping does consider Object/Cursor rorations. Aligning objects would easily be possible without the need to workaround with “Copy All to Selected”. (Without the need to install additional addon).

1 Like

Leonard, your work is amazing! WOW!

1 Like

Hi Davide,
thank you for your positive feedback.

1 Like

Feet …

He still needs mouth cavity and teeth

1 Like

Now the mouth cavity.

Armature used to open mouth.

1 Like

Mouth cavity merged into the whole.

Working on ears …

1 Like

Im finishing topology.

Before I merge mirror, I check the topology flow.
For this purpose I use Subdivisions (Optimal Display disabled) + Wireframe View.
In some areas (like poles) geometry gets ‘crowded’ (agglomerated?). These appear as darker spots. You can see this kind of ‘crowded spots’ on his hips or his shoulder, they are caused by poles.

1 Like

Overall the likeness is definitely there. The topopoly looks nice and clean in the parts you’ve shown. And, your workflow appears to be very disciplined and purposeful. Do you plan to continue with a multires sculpting pass after you’ve finished the retopology?

Yes, there will be a multires pass.

I need Multires to have enough details for hands.
I kept topology of hands as slim as I could. It is enough geometry to have something for armature deformations. Not enough to describe the form, fingers would be too thin after a subdivision without shrinkwrap. Multires is the tool to keep a detailed model, to get rid of the shrinkwrap modifier. But also keep topology of hands slim.

1 Like

Working on cloth.
I am not sure how experienced cloth artists prevent this to happen on cloth simulation in animation in Blender …


Situation …

  1. Cloths are layered.
  2. Cloth layers do pass through other cloth layers.
  3. Aaah … :tophat::horse:munch

Layering cloths is the rule in true life. Except, someone manages to convince everybody to run around in a toga, like old roman did.

Despite layering being common, most information on Blender cloth simulation narrows to a cloth simulation of a single layer.
You do have an idea how to deal with layered cloth simulation in Blender?

1 Like

One approach would be masking the relevant areas of the underlying pieces through mask modifiers. Another approach could be collision modifiers. Each layer could have a collision modifier, but the calculation of the simulation could take a long time and honestly I did never experiment with these layered collisions.

Side notes:
From my experience cloth simulation works best n a larger scale. Real life sizes times 5 ore more tend to give better results. And, you have to experiment with the settings for thickness (outer and inner) for the collision object.

Higher quality steps for the simulation and collision are also required. I usually start from 10 steps.

Before you start the simulation make sure there a no intersections between the different pieces. If there are intersections results are predictably wrong.