Or drop one onto the print. Your image stays on this device.
SVG keeps every stroke editable at any size. Exports always contain the complete drawing.
How the lines find their way
Face contours fits MediaPipe’s 3D facial landmarks to the photograph. It follows the fitted face surface and keeps image flow around it. This works best for a clear, single face.
Surface contours uses Depth Anything V2 Small to estimate relative depth. The lines follow intersections of that surface with tilted planes, so they bend around the nose, cheeks and other volumes. Ink still follows the light in the photograph. Image flow keeps the original edge and texture method.
Depth is inferred from one picture, so fine features and hidden surfaces can be wrong. This is a visible surface estimate, not a complete 3D scan. The example surfaces are precomputed; uploaded photographs are analysed locally with the same models. Models load on first use; no paid inference service is used.
Based on MediaPipe Face Landmarker, Depth Anything V2 and surface hatching research.
Inspired by the hatching in Tri Le’s drawing process. Line direction is inferred from the photograph, so it is an interpretation of the form.