Per-Job Options

Frame rate applied to this job's FootageCaptureData asset and every image sequence linked to it. It always shows the take's current rate and cannot be left blank - editing it writes onto the capture data asset, not just this job. Inert on audio-only jobs, which have no footage.

Start timecode written onto the job's linked image and depth sequences, as HH:MM:SS:FF or HH:MM:SS;FF. Like Frame Rate it edits the CaptureData asset, and is inert on audio-only jobs.

Updates the viewport with each frame as it is processed. Useful to confirm the solver is tracking correctly; it costs a little speed, so leave it off for production batches.

One or more frame ranges the solve skips, e.g. 0-5, 120-125. Use it to cut blinks, obstructions, equipment glare, or any frames where the face is occluded or corrupted.

Per-character expression limits applied to every export this job writes - AnimSequence, Level Sequence Control Rig bake, and FBX. Empty means no limits. The solve itself is unaffected, so Re-Export with an edited profile regenerates every output without re-solving.

Corrects for drift introduced by head movement across the recording. Enabled by default. Disable only if stabilization is degrading a locked-off, tripod-style take.

Enables the tongue tracking solver (applies to Mono and Stereo jobs). Requires UE 5.7 or later. Capture quality must be sufficient for tongue visibility — close-framing and good lighting are prerequisites.

Uses a neutral expression frame to calibrate the facial geometry baseline before solving, with a blend weight (alpha). Available on Mono and Stereo jobs.

Advanced sub-group overriding the models the mono pipeline runs: Smoothing Params (realtime smoothing applied to the solved curves), NNE Backend (inference runtime, e.g. NNERuntimeORTCpu), and the Face Detector, Head Pose Tracker and Face Solver NNE models. Blank or empty on any of them means the engine default - leave them alone unless you are testing a custom model.

The combo in the job settings header, next to Solve. It is what turns body tracking on for a Mono job: Face is the face solve alone, Body the SMPLX body solver alone, Both runs the two together. Body needs the MetaHumanBodyTracker plugin and hips plus ankles in frame; the framing probe picks Both on its own for a full-body take. The Solve Defaults templates still expose the two toggles separately.

Auto-infers the subject's height from footage by default. Disable Auto Body Height to enter a fixed value instead. Epic recommends 145-190cm for best tracking quality.

Threshold for initially acquiring a body lock on the subject. Default 0.7. Lower this if the tracker struggles to acquire lock in cluttered or dim footage.

Threshold for maintaining an existing body lock once acquired. Default 0.5. Lower this if the tracker frequently drops lock during fast motion.

Stabilizes foot contact points to reduce foot sliding during the solve. Enabled by default.

Skips the quick preview pass that runs before the full body solve. Off by default, and only has an effect while Body Tracking is on.

Yaw and pitch limits beyond which a frame is treated as unsolved for head rotation. Defaults: 45deg left/right, 30deg up/down.

What to do with frames outside the allowed head rotation range: None (leave unsolved) or Neutral Pose (substitute a neutral head pose).

Toggle which solved data streams get baked into the exported AnimSequence. Both enabled by default.

Performer Skeleton builds a mesh matching the solve's own body-shape betas. Existing Skeleton retargets the solved body onto a mesh you choose, via an IK Retargeter. Both - the default - runs one pass of each.

Shown only when Body Export Skeleton is Existing Skeleton. Per-job override of the global default target mesh; empty falls back to the global default, then the plugin's SKM_Body.

Shown only when Body Export Skeleton is Existing Skeleton. IK Retargeter used to retarget the solved body onto the Target Mesh. Use the panel's + Create Retargeter... button to auto-generate one pre-wired to the SMPL body-tracker source rig — you only need to configure the target chain mapping in the editor that opens.

Fallback for frames the body tracker couldn't solve: hold the Last Valid Frame, or fall back to a Neutral Pose.

A Skeletal Mesh for face-only jobs, or an Actor Blueprint for body-tracked Mono jobs — MetaHuman Performance requires a Blueprint once body tracking is enabled, since a single mesh can't carry both Face and Body. Use the panel's Create Blueprint... button to auto-build one that leader-poses Face to Body and satisfies MetaHuman Performance's requirements.

Overrides the control rig driving this job's Performance visualization object. Per-job override takes priority over the global default; empty falls back to the global default, then the engine default.

How head movement is represented on the visualization object: Transform Track, Control Rig, or Disabled.

Selects the internal stereo solver algorithm variant. In most productions the default is correct. Change only when directed by Epic support or when experimenting with non-standard rig configurations.

Skips individual solve passes: the preview-quality pass, temporal filtering, the per-vertex solve (expensive — skipped by default for batch), and diagnostic image output.

Override the face contour tracker or face animation solver asset used instead of the pipeline default. Leave empty to use the engine default (shown pre-filled in the settings panel).

The frame number used as the identity reference for head movement correction. Frame 0 is the default. Set to a frame where the subject is in a neutral, stable pose facing forward.

Uses a dedicated neutral expression frame (separate from the performance footage) to calibrate the facial geometry baseline before solving. Improves accuracy when a calibration take is available.

Minimum percentage of the depth map the detected face must cover for a frame to count as valid. Default 80. Frames below it are excluded from the solve; lowering it lets partial-face frames back in.

Minimum face width in depth pixels required for a frame to be included. Acts alongside Face Coverage Threshold to filter out frames where the face is too far from camera or too small for reliable depth data.

Acceptable deviation between the stereo baseline distance recorded in the camera calibration and what is measured in the live footage. Increase if calibration was done at a slightly different rig configuration than the capture.

Acceptable scale mismatch between calibrated and measured stereo geometry. Increase if the rig was adjusted between calibration and capture.

Produces control curves shaped for hand editing after the solve. UE 5.8+, off by default.

Frame used as the Neutral pose. It must have valid RGB and depth and a front-facing view of the performer. Default 0.

Tracks a Teeth pose before auto-rig so the service fits teeth. Off by default - teeth solves are rarely good.

Shown only when Fit Teeth is on. Frame used as the Teeth pose; it must show the performer's teeth clearly, with valid RGB and depth.

Trains the predictive solvers after auto-rig completes, so the Identity is ready for fast performance solving. On by default - stereo takes cannot solve against an unprepared Identity.

Realtime — faster solve, lower quality. Suitable for previews and iteration. Offline — slower, higher quality. Use for final output deliverables.

Emotional baseline the solve colors the animation with: Auto Detect - the default, inferred from the audio - plus Neutral, Happy, Sad, Disgust, Anger, Surprise, Fear, Confident, Excited, Bored, Playful and Confused. It applies to the offline solve; the realtime solve has its own Realtime Mood pair, which defaults to Neutral.

Auto-generates natural blink keyframes when the audio solve does not produce blink data. Enabled by default. Disable if you are driving blinks from a separate system or if the performance data already includes them.

Downmixes multi-channel audio to mono before analysis. Enabled by default. Disable to solve from a single specific channel via Audio Channel.

Which channel from a multi-track audio asset to use when Downmix Channels is off. Defaults to 0 (left/mono). For stereo files where dialogue is on the right channel, set to 1.

Which curves the solve outputs: Full Face (all facial curves) or Mouth Only (lip sync curves only).

Blend weight for the Mood Preset - lower values keep the animation closer to neutral while retaining a hint of the chosen mood. Default 1.0.

Realtime solve only - hidden unless Realtime Audio is on. How far ahead the solver reads incoming audio, trading latency for accuracy. Default 80 ms.