Thursday, August 11th, 2022 Posted by Jim Thacker

Ziva Dynamics launches Ziva Face Trainer

A ‘Ziva Face’ generated by Ziva Face Trainer, running at 60fps at 4K resolution in Unreal Engine 4.

Character simulation specialist Ziva Dyamics has launched Ziva Face Trainer (ZFT), its new machine-learning-trained cloud-based facial rigging service for games and real-time work.

According to Ziva, the process turns a static human or humanoid face mesh into a “high-performance real-time face with full ARKit compatibility” in under two hours.

The firm has also released its ZivaRT plugins, needed to animate the resulting Ziva Face in Maya, or to deploy it in the Unity and Unreal Engine game engines.

Convert any human(oid) head mesh into a real-time puppet capable of over 70,000 facial shapes
Originally known as ZRT Face Trainer, Ziva Face Trainer launched last year in closed beta, before Ziva Dynamics was acquired by Unity.

Built on a 15TB library of 4D scan data, ZFT converts a head mesh uploaded in FBX format into a “real-time puppet” capable of expressing over 72,000 training facial shapes, along with novel face poses.

The resulting Ziva Faces have a memory footprint of 30MB at runtime and “run at 3ms/frame on a single CPU thread”, making them suitable for use in games and real-time applications.

Before a head mesh can be processed, it needs to be cleaned and retopologised manually to match a standard point order, and adequately subdivided to generate convincing deformations.

You can read an overview of the process, for which Ziva Dynamics recommends Wrap, R3DS’s retopology software, and Maya, in this downloadable guide.

You can gauge the results on the product website, which has standard range-of-motion animations for 12 human and humanoid head models.

The real-time puppet can be used in Maya, Unity and Unreal Engine, and has “full compatibility” with Apple’s ARKit framework, used by facial motion capture tools including Epic Game’s free Live Link Face.

Using a Ziva Face in a host application requires the ZivaRT Player software
Users can upload a prepared head mesh to Ziva Face Trainer for free, making it possible to preview the output before buying a training session and downloading the resulting Ziva Face.

To interact with a Ziva Face inside a host application – for example, to set up facial animation in Maya, or to deploy the real-time puppet in a game engine – requires a separate plugin, ZivaRT Player.

The Unity version is free; the Maya and Unreal Engine versions require a paid subscription.

Pricing and system requirements
Ziva Face Trainer is a cloud-based service, so you can upload a head mesh from any standard web browser. ZivaRT Player is available for Maya 2019-2022, Unreal Engine 4 and Unity.

Each training session costs $1,800 and includes one-year subscriptions for ZivaRT Player for Maya and UE4. Additional subscriptions for either edition cost $200/year. ZivaRT Player for Unity is free.

Read more about Ziva Face Trainer on Ziva Dynamics’ website

Read more about how to prepare a face mesh for Ziva Face Trainer in the online documentation