From install to first take
Eight panels, in order — each one unlocks the next.
- 1
Install & set up
Install the zip via Edit > Preferences > Add-ons, then enter your licence key. Open the Face Addon tab in the viewport sidebar (N).
- 2
Get a character fast
No rig of your own? Add Character drops in a ready-to-go pre-rigged, pre-weighted, pre-mapped head — mesh, rig, or both.
- 3
Prep landmarks
Auto-Place Landmarks builds the correspondence map hair, pinning, and identity fitting depend on. Fit Identity or manually pin to refine the fit.
- 4
Generate & map shape keys
Generate Shape Keys if needed, then Auto-Detect Mapping to match landmarks to your shape keys. Save the setup as a preset.
- 5
Map bones
For armature-driven characters: Auto-Detect bone mapping, Map Head for head rotation/position, then manual mapping for anything left over.
- 6
Puppet from video
Analyse Video, run auto-mapping, solve bones if needed, then Bake to Keyframes to get editable F-curves.
- 7
Puppet live
Pick a camera, Start Live, Set Neutral Pose to calibrate, then perform — Record captures the take to keyframes.
- 8
Add hair
Match Hair to Photos or pick a preset, then Add/Update Hair to generate strands. Bake Hair to Mesh to export.
Hardware requirements
The tracking and the solve happen on the Expression Server, so the add-on asks far less of your setup than face capture usually does. What you need comes down to which of the two paths you use.
Offline video
You hand the add-on a video file. The Expression Server does the tracking and the solve, and sends back finished bone transforms ready to apply to your rig.
What you need: Blender, and a connection to upload with. If Blender runs well for you, this path does too.
Live
A camera feed is tracked frame by frame and driven onto your rig as you move, so you see the character react while you perform.
What you need: real CPU headroom. Live tracking runs continuously at 30fps while Blender is also running, so this is the demanding path — size your machine for it, not for the video workflow.
Coming soon — not yet enabled in the current release.
Blender
- Minimum
- 4.0
- Recommended
- 4.2 LTS or newer
Enforced by the add-on manifest — Blender itself refuses to install it below 4.0. Our test suite also passes on 3.2, but that is not a supported configuration.
Operating system
- Minimum
- Windows 10, macOS 12, or Linux
- Recommended
- Windows 11
The add-on is pure Python against Blender’s own API with no platform-specific code, so anywhere Blender runs, it runs. Windows is the most heavily exercised.
CPU
- Minimum
- 4 cores — Blender’s own floor
- Recommended
- 8 cores, for Live
Live tracking is continuous and CPU-bound, and it shares your machine with Blender — 4 cores will run it, 8 will keep it smooth. The video workflow is undemanding by comparison: the solve is not your CPU’s problem there.
Graphics
- Minimum
- Anything Blender supports
- Recommended
- —
No GPU is used for compute, on either path. No CUDA, no dedicated card, no VRAM floor. Your GPU matters exactly as much as it does for ordinary Blender work and no more.
Memory
- Minimum
- 8 GB — Blender’s own floor
- Recommended
- 16 GB
The add-on adds the cost of your own scene rather than a fixed model footprint. We have not published a measured peak figure because we have not measured one; treat these as Blender’s numbers, not ours.
Disk
- Minimum
- A few MB, plus your project
- Recommended
- —
The add-on is a small zip and downloads nothing at runtime — no model weights are fetched on first use. Your .blend files and source footage dominate.
Network
- Minimum
- Broadband
- Recommended
- Fast upstream
Your footage is uploaded for solving, so upload speed sets how long you wait on long clips.
Camera
- Minimum
- Any webcam, or a phone running an IP-camera app
- Recommended
- 720p at 30 fps
Live only — the offline path needs no camera at all. Higher resolution costs you frame budget rather than buying accuracy, so 720p is the sweet spot.
| Component | Minimum | Recommended |
|---|---|---|
Blender | 4.0 | 4.2 LTS or newer Enforced by the add-on manifest — Blender itself refuses to install it below 4.0. Our test suite also passes on 3.2, but that is not a supported configuration. |
Operating system | Windows 10, macOS 12, or Linux | Windows 11 The add-on is pure Python against Blender’s own API with no platform-specific code, so anywhere Blender runs, it runs. Windows is the most heavily exercised. |
CPU | 4 cores — Blender’s own floor | 8 cores, for Live Live tracking is continuous and CPU-bound, and it shares your machine with Blender — 4 cores will run it, 8 will keep it smooth. The video workflow is undemanding by comparison: the solve is not your CPU’s problem there. |
Graphics | Anything Blender supports | — No GPU is used for compute, on either path. No CUDA, no dedicated card, no VRAM floor. Your GPU matters exactly as much as it does for ordinary Blender work and no more. |
Memory | 8 GB — Blender’s own floor | 16 GB The add-on adds the cost of your own scene rather than a fixed model footprint. We have not published a measured peak figure because we have not measured one; treat these as Blender’s numbers, not ours. |
Disk | A few MB, plus your project | — The add-on is a small zip and downloads nothing at runtime — no model weights are fetched on first use. Your .blend files and source footage dominate. |
Network | Broadband | Fast upstream Your footage is uploaded for solving, so upload speed sets how long you wait on long clips. |
Camera | Any webcam, or a phone running an IP-camera app | 720p at 30 fps Live only — the offline path needs no camera at all. Higher resolution costs you frame budget rather than buying accuracy, so 720p is the sweet spot. |