How to turn a 2D picture into a 3D mesh in Cympha Studio

Face from Picture builds a 3D face mesh from one portrait photo in Cympha Studio, standalone or registered to a CBCT, so the dental case can include a face without a facial scan.

Audience: Clinicians, case reviewers, and designers who have a single 2D portrait photo of the patient and want to bring it into Studio as real 3D geometry — either standalone or registered to the patient's CBCT.

You'll come away knowing: which Studio action turns a front-facing photo into a 3D face mesh, when to use the CBCT-aligned path versus the standalone (no-CBCT) path, what the on-screen sliders control, and how to land the result in the same scene as the rest of the case.

Time required: A few minutes for a face. Slightly longer the first time, when the Head mode fetches a one-time model download (cached by your browser after that).

Before you start

You need:

A single front-facing portrait photo of the patient (.png, .jpg, .jpeg, or .webp). Most of the face should be visible, eyes roughly level, mouth either closed or in a natural smile. A photo cropped tightly to the head works fine — Studio re-crops internally for the AI step.
Cympha Studio open at studio.cympha.com.
Optional — a CBCT volume of the same patient already loaded in the scene if you want the reconstructed face to register against bone (this is the "with CBCT" path below). See Load a CBCT or CT in Cympha Studio for loading instructions.

A note on what this feature is, and isn't. Studio's 2D-to-3D action is a face reconstruction — it estimates a 3D face mesh from one portrait photo using on-device AI (facial landmarks for the Face mode, monocular depth for the Head mode). It is not a generic "image to mesh" tool for arbitrary objects, and it is not a way to turn a lateral cephalometric X-ray into 3D — for that, see Cephalometric tracing. It is also distinct from importing a facial scan captured by a dedicated 3D scanner — that workflow is documented in Load a facial scan and work with the 3D camera.

Step 1 — Open Face from Picture

In the left tools dock (Facial Scan toolbox), click Face from Picture. A file picker opens — choose the patient's portrait photo. (If your scene is empty and you'd rather start straight from the welcome screen, the same action is available under the welcome card.)

Studio loads the photo and decides which path to use next based on whether a CBCT is currently in the scene.

Step 2A — Without CBCT (photo only)

If no CBCT is loaded, Studio goes straight to the 3D Face Reconstruction panel. You'll see your photo as a billboard in the viewer with the panel title "3D Face Reconstruction" on the right.

Pick a mode:

Face — a tight mesh of the face only (forehead, eyes, nose, cheeks, lips, chin). Built from facial landmarks detected on the device — no large download. Fast and clean for smile-design and chairside communication.
Head — a larger mesh that includes hair, ears, and neck. Built from monocular depth. Takes longer the first time (a one-time model download), then runs quickly from the browser cache.

Adjust as needed:

Brightness / Contrast — re-tone the photo before reconstruction.
Resolution — triangle density of the mesh (release the slider to re-run with the new value).
Nose Depth — only shown on the no-CBCT path. Without a CBCT to lock metric depth, nose projection is the most variable dimension; this slider lets you push the nose in or out until the profile looks right.
Face Box — when shown, expands the cropped region the AI sees.

When you like the result click Done to keep the mesh in the scene, or Cancel to discard.

Step 2B — With CBCT (registered to bone)

If a CBCT is already loaded when you click Face from Picture, Studio inserts two extra steps so the resulting face mesh ends up in the same coordinate frame as the CBCT.

CBCT Alignment. The "3D Face Reconstruction — CBCT Alignment" panel opens. Drag the HU Threshold slider until the soft-tissue surface (skin) is visible in the 3D view, with the head in portrait orientation. Click Use DICOM to detect CBCT landmarks and align to them, or Skip to fall back to the no-CBCT pipeline. Cancel aborts.
Landmark Verification. The "Landmark Verification" panel opens with the photo over the CBCT. Click Face Detection to detect facial landmarks on the photo, then verify they sit where they should. Use Opacity to fade between photo and CBCT, and (when shown) Face Box to expand the region considered. Click OK when the landmarks look right.
Reconstruction. The 3D Face Reconstruction panel opens with Face and Head modes, same as the no-CBCT path — except Nose Depth is hidden, because the CBCT now anchors the depth scale. Pick a mode, adjust Resolution, click Done.

The reconstructed face arrives already aligned with the CBCT and intraoral scans. There is no separate registration step.

What happens next

The reconstructed face becomes a regular mesh in the Model browser alongside any intraoral scans, CBCT, or implant plans. You can:

Hide / show it, recolor it, or remove it like any other model.
Use the orthographic camera presets (front / patient right / patient left) to take repeatable screenshots — see Load a facial scan and work with the 3D camera for the camera workflow.
Save the whole scene with ☰ → Save project so the reconstructed face travels in the .cyp project file with the case.
Capture it into the Media gallery using the camera icon on the top toolbar; the picture lands next to your other case media.

For a fuller scene assembly across CBCT + facial reconstruction + dental scans, see Build a virtual patient in Cympha Studio.

Troubleshooting

"No face detected in photo." The face detector couldn't find a face. Most common causes: the head is rotated too far away from the camera, the face is too small in the frame, or the photo is heavily filtered. Cancel, re-crop or replace the photo, and try again.

Head mode takes a very long time on the first run. The Head mode fetches a one-time model download per browser. After the first reconstruction it is cached locally and subsequent runs start quickly.

Nose looks too flat or too pointy (no-CBCT only). Slide Nose Depth up or down — the dialog re-runs the reconstruction when you release the slider. Without a CBCT to lock real-world depth, this is expected variability.

The face mesh isn't aligned with my CBCT. You probably went through the no-CBCT path. Either the CBCT was loaded after you started Face from Picture, or you clicked Skip in the CBCT Alignment panel. Cancel, make sure the CBCT is visible in the scene, and click Face from Picture again — the CBCT Alignment panel should appear first.

I want to turn a lateral X-ray into 3D. This feature reconstructs from a portrait photo, not from an X-ray. For 2D X-ray work, use cephalometric tracing — Studio traces landmarks on the X-ray but does not generate a 3D mesh from it.

Last updated 2026-05-30.