Avatar SDK UE plugin  2.0.0
Realistic avatar generation toolset for Unreal Engine
Main Page


Meet the Avatar SDK: robust AI-powered 3D avatar creation engine!

Useful links:

Available pipelines

We are constantly working on our avatars' improvements in quality, recognizability, and feature set. This way we have released several avatar pipelines. Each avatar pipeline may have a few pipeline subtypes (different sets of available features, different model topologies, etc). You can find a list of avatar pipelines with their subtypes that are currently supported in the UE plugin below:

  • Head 2.0 | head/mobile (head_2.0 pipeline, head/mobile subtype)
    Predicted head with a detachable haircut and set of 51 facial blendshapes. The mesh topology of a head is always the same.
  • Head 2.0 | bust/mobile (head_2.0 pipeline, bust/mobile subtype)
    Predicted bust with a detachable haircut and set of 51 facial blendshapes. Unique mesh topology.
  • Body 0.3 | mobile (body_0.3 pipeline, mobile subtype)
    Predicted bald head and full body model with fixed topology, predicted and artificial haircuts, and artificial outfits. Predictions of the body shape may be improved by specifying the height, weight, gender, and body measurements of the model. A model skeleton is also available for use with animations. Please note: the pipeline supports adult bodies only. If you specify parameters as for a child (small height and small weight) then the pipeline will return a body following these parameters, but it will be the body of a small adult, not a child.

Avatar SDK Cloud API supports different LODs for Head 2.0 Avatars. We plan to add support for Full Body avatar LODs in the next version.

These pipelines are available on the Pro subscription plan (Avatar SDK pricing) only!

Avatar SDK pipelines

Supported platforms

The current version of the Avatar SDK UE Plugin is supported on Windows and macOS. Version 1.0.2 of the plugin is supported on Windows, macOS, iOS, and Android and does not support Full Body avatars. Please, refer to Platform specific information if you need more details.

Supported Unreal Engine versions

Version 2.0.0 of Avatar SDK Plugin is supported for Unreal Engine 4.25 and 4.26. The previous version of the plugin (1.0.2) is supported for Unreal Engine 4.24 and 4.25 and does not support Full Body avatars.

Getting Started

Please follow the instructions on the Getting started page.


Is avatar generation limited?

Yes, the limit of the avatar generation depends on your subscription plan: Avatar SDK pricing

Does it work in runtime?

Yes, the UE Plugin works in runtime. Any user of your application can generate his or her avatar.

How many triangles do the avatars have?

8 LODs supported ranging from ~4K to ~24K triangles per avatar for the head_2.0 pipeline, head/mobile subtype:

  • LOD0: corresponds to the base head mesh, contains 11894 vertices and 23404 triangles.
  • LOD1: 9969 vertices, 19054 triangles.
  • LOD2: 8732 vertices, 16860 triangles.
  • LOD3: 5713 vertices, 11007 triangles.
  • LOD4: 4953 vertices, 9497 triangles.
  • LOD5: 4259 vertices, 7889 triangles.
  • LOD6: 3789 vertices, 6937 triangles.
  • LOD7: 2034 vertices, 3926 triangles.

head_2.0 pipeline, bust/mobile subtype has ~50K triangles per avatars. There are no LODs for this subtype.
The generated haircut of the head_2.0 pipeline has a configurable number of polygons.

How can I export the generated model?

Models export feature will be implemented in future releases of Avatar SDK UE Plugin.
It is still possible to access downloaded files: in the sample scene, their location is defined by the GetDefaultAvatarDestination function.

Can I display the avatars of the different users in my app?

Yes, you can query any generated avatar if you know the avatar code.

What recommendations can I give my users for input photos?

The SDK is pretty robust, but these recommendations will improve the quality (in priority order):

  • The face should be easily recognizable in the photo. The photo should not be blurry.
  • Uniform good lighting on the photo, without dark shadows or too bright glare.
  • A person should keep a neutral facial expression or slight smile but without opening a mouth. If teeth are visible on the input photo the lip texture might be incorrect.
  • It's best to look straight into the camera without turning your neck or eyes.
  • We advise removing the glasses because they're reconstructed only in the texture, not 3D mesh. But this is not completely necessary.