Capture guidelines
How to take the photos for a head: what to shoot, how to light it and what to avoid.
What to shoot
For Scan Reconstruction: 2–15 photos of one person, from different angles.
- Always include a straight-on view, plus turned views to the left and right (~40°).
- Keep the face neutral, centred and fully visible. Tie back long hair and show the ears.
- Shoot in even, indirect light, and keep the subject still.
- Ask someone else to take the photos while you stay still.
- Human faces only.
What can hurt your result
- Face isn't sharp
- Fine detail is what the geometry is built from. Keep the camera steady and the subject still.
- Face is too small
- There isn't enough detail to reconstruct from. Move closer, or crop the photo so the face fills more of the frame.
- Face is cut off
- Part of the face runs past the edge of the photo. The missing area has to be guessed, which can leave artifacts in the result.
- More than one face
- We can only build one head at a time, and we may pick the wrong person. Use a photo with only the intended person in it.
- Expression isn't neutral
- Smiling or talking changes the face shape. A relaxed, neutral face reconstructs best.
- Eyes look closed
- Closed eyes distort the eye region of the model.
- Glasses detected
- Frames and lens reflections hide the shape of the eyes and temples, and that distorts the head we build. Retake without glasses if possible.
- Something may be covering the head
- The shape of the head is what we're building, so anything covering it — a hat or hood — can distort the result. An uncovered head works best where possible.
- Head is tilted
- Keep the head upright rather than tilted to one side.
- Photo is too dark
- Detail is lost in shadow. Face a window or turn a light on.
- Photo is too bright
- Highlights are blown out, so surface detail is lost. Avoid direct sun or flash straight at the face.
Scan Reconstruction
A head and matched cameras from prepared photos.
Scan Reconstruction reconstructs head geometry and texture directly from your photo set, so the photos are used as given. Capture the subject from multiple angles with even lighting and a neutral expression, ideally using a synchronized multi-camera rig.
You get a head in MetaHuman topology, plus a 3D scene with solved camera positions and focal lengths for use in DCC tools. With Scan Reconstruction off, Studio works from casual photos or images from the web instead.