Make an image to depth map conversion you can trust

Image-to-depth estimation reads a single view, so the result is scene-relative rather than measured. Transparent objects, fine hair, and heavy shadows can confuse any estimator. A useful result preserves the shapes that control the next operation while suppressing background distance, edge halos, and texture that would become false geometry. DepthCarve keeps the source beside the estimate and exposes range controls before export, so a technically valid file does not get mistaken for production-ready input.

Choose a sharp source with a clear subject and quiet background. Crop decorative borders and distant scenery before generation so they do not consume the useful range. Black point removes low-value range, white point sets the upper boundary, contrast separates meaningful levels, and inversion chooses which end becomes raised or near. These controls are applied to the output instead of triggering another inference, which makes comparison fast and keeps model usage predictable.

Start with the free 8-bit PNG, compatible 16-bit container, and low-resolution preview STL to confirm orientation, crop, silhouette, dimensions, and import behavior in LightBurn, CAM packages, slicers, or compositing tools. The compatibility PNG is explicitly preview-derived; it is not sold as added precision.

Approve one production job when raw float depth and final files are justified. One ZIP contains forward and inverse true 16-bit maps, preview, watertight STL, raw EXR, and a parameter and validation manifest. Provider normals, point cloud, GLB, and camera links remain temporary and are recorded for immediate retrieval.

An 8-bit grayscale preview is for checks and first passes. It cannot carry metric truth, and it does not replace a calibrated measurement in the target application. Zoom into boundaries, look for isolated bright or dark islands, and test the output at the final aspect ratio. A higher mesh grid or a 16-bit container cannot repair an incorrect scene interpretation. Correct the crop, source lighting, or depth range first; only add density after the underlying field is useful.

DepthCarve makes source checks, editing, 3D preview, manufacturing checks, and compatibility exports free because approval should happen before a provider job is charged. A provider failure refunds the original one-credit debit with a unique reference. That protection covers technical failure, not an aesthetically unsuitable completed estimate.

The browser compresses the selected image before it leaves the device. The source is sent to the chosen inference provider only when Generate is pressed, and D1 stores job state, an input fingerprint, and credit entries rather than the original file. Model output URLs can expire, so download approved assets into the normal project archive instead of treating a provider link as storage.

Every credit movement carries a unique reference. Repeated callbacks cannot grant the same purchase twice, the same source-job key cannot create duplicate provider work, and rebuilding an available bundle does not keep charging. External requests have bounded attempts and timeouts. Clear failures remain visible instead of hiding behind a permanent loading state.