Make Lithophanes from Photos

Overview

You have one photograph on your desk, a lamp base on the shelf, and one goal: a backlit panel that shows the portrait. The stuck point is the middle, where color has to become thickness. One repeatable chain closes that gap: prepare the photo, generate a relief-type 3D model, repair the mesh, export at size, verify the slice, test print.

V2Fun's browser-based image-to-3D workflow covers the generation stage, starting from one reference image, and the result is a lithophane base whose light and dark live in physical thickness, because the print file holds no color. Lithophane-specific numbers, translucent wall values among them, are marked as a knowledge gap; specialized tools own that layer. Run it all on your own slicer and printer.

From a photo to a printable lithophane base

A lithophane is a thin relief panel that reveals its image when lit from behind. Your first move needs no software: measure the photo against the published input guidelines, which ask for a clear JPG, PNG, or WebP image of a single subject, adequately lit, on a plain contrasting background, with an angle that shows front, side, and depth cues and a longest edge of at least 1,000 pixels.

Glare, motion blur, heavy shadows, cropped edges, and overlapping objects should be fixed or retaken first, because they travel straight into the model. Relief-type objects sit in the documented single-image-start group; the multi-view-recommended list covers full-body figures, product shells, jewelry, layered clothing, and back-detail-critical objects. Relief work starts from one photo, and depth loss or flattening still needs checking afterward.

Cleanup help exists for imperfect photos: the AI Image page describes the generation and reference-image tools for that preparation step.

Check missing details and scale errors

Two edge failures recur. Hidden, cropped, or blurry photo details cannot be reliably reconstructed at generation. A file imported at the wrong scale through unit ambiguity can pass a visual glance while being wrong in every dimension.

Scale first, then judge walls: a feature that survives at 200 mm "may disappear or break at 30 mm." That single habit prevents the most expensive mistakes in this chain.

The prerequisite pass

Before any generation, decide the print job itself. The documented prerequisite asks seven questions: the purpose of the object, its target physical size, the printer process and material it will run on, the level of detail it needs, its minimum strength requirement, how sensitive it is to supports and surface finish, and whether color or material data matters for the finished part. Your answers set the wall-thickness range and the acceptance bar for every later step.

Identify the stage responsible for a defect

When the finished part is wrong, identify the responsible stage first. A missing back points to generation. Disappeared small features point to scaling or the printer's resolution limit. Heavy scarring from supports is an orientation problem. As the documentation puts it, "Identifying the responsible stage prevents unnecessary editing."

For a deeper walkthrough, the operational tutorial on the official blog covers preparation, generation, retopology, export, and the slice check in order. Export is not acceptance yet.

Verification happens in the slicer

Five checks make up the documented acceptance pass: scale, watertight mesh, wall thickness, stability, and detail level. The reasoning compresses to one line, "The unsliced 3D view is not enough." The layer preview shows what the 3D view hides, including unsupported islands, features too fine for the printer's resolution, and a weak first layer.

Verify units and print formats

Unit errors are common and easy to miss. STL carries no standardized unit declaration, so different software can read the same file at different scales, and the documented practice is a numeric bounding-box check after import. Binary STL in millimeters, the tested configuration, reduces the ambiguity.

The 3MF-versus-STL choice runs through five documented questions, from unit declaration and component metadata to whether color and material data travel with the file. Two lines settle the rest: "Always inspect the imported dimensions and geometry," and "A better format can reduce ambiguity, but it cannot guarantee printability."

Process choices and habits

On process, the documented comparison stays generic: resin reproduces detail noticeably better than FDM, while FDM shows visible layer lines and costs less. Translucent wall values, backing plate parameters, and resin plate settings fall under the knowledge gap and belong to specialized tools and community references.

The top habit is to set real-world size first, then judge walls, because thickness only means something at final dimensions. The next habit is to run the five-check pass and the layer preview before any full-size print.

Where V2Fun speeds up the photo stage

V2Fun's help concentrates on the slowest link in this chain: the jump from photo to relief draft. Browser-based image-to-3D generation turns one JPG, PNG, or WebP reference image into a relief-type draft previewable through a full 360 degrees, with no local software. Mesh correction, real-world size, and slice verification stay with the tools that own them.

Upload the photo once, and inspect the full 360-degree turn, because that preview is where depth loss and flattening surface on single-image starts. The check matters most for lithophanes: thin walls begin as thin features in this draft. The current official snapshot records GLB and 3MF as the confirmed export pair; STL appears in V2Fun's blog tutorials, a claim below a confirmed menu entry, so the export menu is the final authority. Generation consumes credits, so check the pricing page for current rates before a batch.

On preparation, the official wording is that "V2Fun can also help at the preparation stage," with the three listed capabilities being "AI image generation, partial repainting, and smart reference-image generation." The concrete move: upload the photo, run image-to-3D generation, inspect the full turn, and export at real-world size before the slicer's five checks.