Pic23D Open converter

About Pic23D

A small browser-side converter built around one rule: the file you download should be printable at the size you asked for, with nothing in between.

What Pic23D is

A single-purpose tool: take an image, turn it into a printable solid, give you the file in millimetres. Two families of tools — the height-map converters (lithophane for photographs, relief for grayscale height maps, plaque for logos and cut-outs, exported as STL or OBJ) and the AI generator, which reconstructs a full 3D model from one photo.

Why the converters run in the browser

The conversion is arithmetic: sample a grid, map brightness or transparency to thickness, triangulate the grid, cap it into a closed solid. That job fits in a browser tab, and doing it there means no upload, no queue, no account and no waiting. It also means the tool costs almost nothing to run, which is why there is no paywall, no credit counter and no watermark on your download.

Why the AI mode does not

Reconstructing a 3D object from a single photograph needs a large model that cannot run on a phone, so that mode is the one feature here that sends your image to a server. It goes to our worker, then to Tencent's Hunyuan 3D API, and the resulting mesh comes back to the browser; we do not keep the image or the model afterwards. Because that call costs real money per generation, the AI mode is limited to a small number of free runs per person per day, while the browser converters stay unlimited. The free page spells out exactly which parts are limited.

What it deliberately does not do

The height-map tools do not reconstruct 3D objects from photographs. Turning one photo into a full 3D figure needs a model trained on 3D data, and the honest description of that output is "a guess" — which is why that job lives in the AI mode, with its own limits. If your job needs a reconstruction, the browser converter is the wrong tool and the 2D to 3D guide says so in the first paragraph.

What the browser gives you is the thing a hobby printer can actually deliver tonight: a watertight relief at the exact size you asked for, printed on one colour of filament.

How the site is put together

Everything except the AI mode is static HTML and vanilla JavaScript. The mesh builder writes binary STL directly into a memory buffer — no libraries, no uploads — and the OBJ writer deduplicates vertices before emitting faces, which keeps file sizes reasonable. The AI mode is a single Cloudflare Worker that signs requests to the Tencent Cloud API; the browser never sees an API key. Both modes share one WebGL viewer so you can rotate the result before you download it.

Who it is for

People with a 3D printer and a photo on their phone: keychains of a shop logo, a lithophane of a grandparent, a terrain relief from a height map, a sign for a workshop door. If that is your job, the whole workflow takes about two minutes.

Contact

Questions, bug reports, or a request to remove a page: hi@pic23d.com. There is no account system, so we cannot look up an individual visitor — but email reaches a person.

Questions

Why is there no upload and no account?

Because the conversion needs no server — decoding an image and triangulating a grid is fast enough in JavaScript, so there is no reason to make you upload a photo and wait.

Do you keep the files I make?

No. The meshes are generated by you, in your browser, from your own images. Nothing is collected, resold or trained on.

Is the tool free?

Free to use, with no watermark and no limit on how many files you build. The detail range is capped at 300 points across so a phone browser does not run out of memory.