The Virtual Form

THE VIRTUAL FORM

Modeling and Sculpting the Face

0. Introduction

1. 3D Scanning using Photogrammetry
2. Setting up the References
3. Base Mesh Modeling
3b. Base Mesh Modeling Using Scan Retopology
4. Subdividing and Tweaking the Mesh
5. Modeling and Stitching the Ears
6. Texturing using Photographs

7. Refining the Face Using Digital Sculpting
8. Adding Details & Asymmetry
9. Texturing Using Polypainting
10. Adding Hair
11. Making a Watertight Form
12. Rendering
13. 3D Printing

1. 3D Scanning using Photogrammetry

Software : Agisoft Photoscan 1.1.6
Hardware : Canon 5D Mark III 50mm 1.2 (f 5.6 - 1/160sec)
Setup Playback Speed : x2
Processing Playback : Skipped

Using photogrammetry on an unstable target (in this case the head of the model) with a single camera can be quite difficult. Modern photogrammetry setups often use more than 50 cameras at once and there is no limit to that number, the more cameras the better the scan can be. Here we are not aiming for a perfect scan but for a rough one that we can use in 2 ways, first, to create some orthographic backdrops and then, to keep it as an extra reference.
The scan was done using 21 photos. It’s important that the camera moves and not the model. Everything should be taken in manual modes ; the aperture, shutter speed and ISO should remain the same. In this case, only natural light was used. The fluorescent lights should be avoided ; its high speed flickering makes the pictures’s light condition inconsistent.

2. Setting up the References

Software : The Foundry MODO 801 SP5
Playback Speed : x2

A photo taken with a “classic” camera has inevitably perspective in it. With perspective, the closer an object is, the bigger this object appears. In a front portrait for example, the nose will always appear bigger, relatively to the ears. Having a 3D scan we can create some orthographic views. With an orthographic projection, which doesn’t contain any perspective, everything is shown at the same size. Therefore, this type of images makes perfect references. Also, the information of the front view will match exactly the information of the side view. It wouldn’t be the case if they had perspective.
Here, after positioning the scan, we render the 2 views, reimport them and set them up as backdrops. We end up with both 2D and 3D references.