Virtual Human & Motion Capture Production
A virtual performer built for a live broadcast segment: four days of optical motion capture, 90 facial solves and a real-time Unreal Engine 5 stage rig driven by two operators. Astel Games delivered the character, the mocap cleanup and a playback system that ran three hours without a restart.
- Client
- Riot Games
- Engine
- Unreal Engine 5
- Platforms
- PC, Broadcast
- Year
- 2023
- Duration
- 6 months
- Team size
- 11 specialists
- Disciplines
- Motion capture, Facial & performance animation, Real-time virtual human, Look development
- Category
- Virtual humans & mocap
Artwork from our archive. High-resolution originals are available on request under NDA.
What the client needed.
Facial capture ran on a performer wearing a head-mounted camera under coloured stage lighting, which introduced tracking noise around the jaw and eyelids. The broadcast also required a clean switch between the live performer and a pre-recorded take without dropping the character.
- Calibrated the head-mounted camera rig daily and stored a lighting-specific solve baseline for the stage.
- Cleaned 90 facial takes by hand in Maya, fixing jaw and eyelid noise frame by frame where solvers failed.
- Built a two-machine Unreal setup with a hot spare sharing one Take Recorder feed.
- Rehearsed the live-to-recorded switch 14 times in the venue before broadcast day.
- Prepared a baked animation fallback for every segment, loadable in under 20 seconds.
- The full three-hour broadcast ran with no character restarts and no visible tracking pops.
- 90 facial solves were delivered in five weeks against an initial eight-week estimate.
- The stage rig and rollback procedure were documented for the client's in-house team.
Have a similar production coming up?
Tell us the scope and the date. We will tell you what it takes — honestly, including what we would not promise.