Developers: Integrate FaceGen technology into your software:

  • Main SDK:
    • Create faces with age, race, gender, shape and color controls.
    • Create an infinity of random, realistic faces at run-time.
    • Load faces created using FaceGen Modeller Free version.
    • Code faces in 130 bytes for low bandwidth or compressed storage applications.
    • Integrate FaceGen into your content creation pipeline.
    • Full C++ source code.
  • Full SDK (Main SDK + PhotoFit):
    • Create a 3D face from one or more photos.
    • Works in combination with all Main SDK functionality.
    • C API dynamicially loaded libraries for multi-platform.

Overview
Design
Integration
Competitive Analysis
More Information

Overview

FaceGen SDKs have been licensed by game developers, universities, movie studios, and companies selling 3D authoring tools, web avatars, face recognition software and facial surgery visualization services.

The Full SDK can turn a properly-taken photo of a face into a FaceGen-coded face in about 5 seoconds per photo on a modern multi-core PC.

The Main SDK gives your end users all of the face controls seen in FaceGen Modeller so they can create and edit their own faces from scratch or edit faces from the PhotoFit.

Age Morphing

The Main SDK can also be used to create a face stored in an FG file. Since the free version of Modeller allows saving of FG files, your end users can use the full power of FaceGen Modeller on their PC then import their face into your application as an FG file. Your application retains full control over all mesh topologies and accessory models; the FG file only specifies the face shape and colouring.

Either way, the Main SDK applies these faces to any head mesh. Use one of ours, or integrate your own using FaceGen Customizer.

The Customizer also allows you to add in you accessory models such as hair or glasses, and have them automatically fit seamlessly with any face. Or apply your faces to fantasy races. The Customizer is included with the SDK.

FaceGen is carefully designed so that animation morphs will automatically conform to any face you create.

Design

Faces are created or modified according to the various parametric controls, and represented by a 130-number coordinate vector in "face space" (these are the same numbers stored in a FaceGen FG file). In the Controls step, the face is manipulated in face space:

In the Coding step, a face space coordinate is used to construct a 3D textured model:

The coding step is repeated for every model which must fit seamlessly with the face; a hairstyle and glasses for instance.

Dataflow

Integration

The output of the FaceGen Coding stage is an updated list of vertex positions and, if applicable, a new texture image.

FaceGen shape statistics always leave the neck seam unchanged for seamless stitching to the body.

A skin colour texture can be generated for the torso or body, in order to create a seamless match.

If you prefer to put your hair, helmet, etc. into the head texture rather than a separate model, it can be composited directly onto FaceGen's skin-coloured output.

The vertex positions output from FaceGen can be used to update any kind of positional information. For instance, vertices can be used to define bone positions as well as polygonal vertices. Once the new vertex positions have been computed with FaceGen, they can be applied to your proprietary structure (polygonal, subdivision, bones, etc.) since the vertex ordering remains unchanged.

Competitive Analysis

Main SDK

The statistically derived face shape and coloring controls in the Main SDK are unique in the market. Here is a comparison with artistically created controls:

  • Pros:
    • Greater realism.
    • Many innovative face-editing controls:
      • Click-and-drag freeform shape edit
      • Genetic interface
      • Age, race, gender and 150 other parameters
      • Random face creation with or without constraints
      • Caricature and asymmetry parameters.
      • Code available for Windows, Mac OS/X and Ubuntu.
    • Can be applied to custom mesh topologies and UV layouts.
  • Cons:
    • Greater computational requirements.
    • Lack of very small shape details such as nose tip shape and wrinkles (these can be added by traditional means).

Full SDK (Main SDK + PhotoFit)

Our approach of fitting a 3D statistical model to a photograph is unique in the market. Other solutions use photo mapping, in which the final texture is a composite of the photographs. Here's how our approach compares:

  • Pros:
    • Greater realism when seen from angles different than the photograph(s).
    • Can be combined with our innovative face-editing controls.
    • Can be applied to custom mesh topologies and UV layouts.
    • Creates a bald head which can be used with any hairstyle.
    • Very low bandwidth coding of unique shape and texture.
    • Multithreaded code available for Windows, Mac OS/X and Ubuntu.
  • Cons:
    • Cannot use the hair in the photograph.

More Information

Email us for Documentation, API and pricing.


Home | Products | Downloads | Contact | About SI | Support

© Singular Inversions Inc.