Website powered by

Low Budget Facial Capture Pipeline

Facial Capture Featurette | Letters From Vega Behind The Scenes

Crandall in engine during the interview scene.

Crandall in engine during the interview scene.

The Crandall facial rig. Each set uses 51 blend shapes to drive the expressions. A number of which use driven keys to drive the jaw and the eyes.

The Crandall facial rig. Each set uses 51 blend shapes to drive the expressions. A number of which use driven keys to drive the jaw and the eyes.

The Dalynn facial rig in Maya showing the animation for e1s4a_L10 (final line in the scene).

The Dalynn facial rig in Maya showing the animation for e1s4a_L10 (final line in the scene).

The Dalynn  rig in the CryEngine character tool showing the animation for e1s4a_L10 (final line in the scene).

The Dalynn rig in the CryEngine character tool showing the animation for e1s4a_L10 (final line in the scene).

The final Dalynn character in CryEngine showing the animation for e1s4a_L10 (final line in the scene)

The final Dalynn character in CryEngine showing the animation for e1s4a_L10 (final line in the scene)

Low Budget Facial Capture Pipeline

Due to the requirement in Letters From Vega for custom facial animations a pipeline on a low budget was developed in July 2017. Faceshift was used with a Microsoft Kinect V1 for the capture with Maya used for the re-targeting and cleanup section of the pipeline. In total three rigs were created (2 male, one female) for Letters From Vega with the two male rigs created with compatibility with CiG assets from Star Citizen in mind. This compatibility would allow simple animations for other head rigs without the need to re-target the animations. This would give poor quality movement up close but adequate movement at a distance. This setup allows portability of recording software due to the cheap price of a kinect sensor and the ability to transfer and load recordings between the Voice Actor and animation artist.

For consistency throughout the pipeline all lines of dialogue are captured separately and given a designation based on a label in the script.