Sketch or AI Render to a Castable CAD File (2026)

Search for a way to turn a sketch or an AI render into a file you can cast, and most of what comes back is a person. Freelancers on Etsy and CAD bureaus, quoting per piece, turning your drawing into geometry by hand. That is still the most reliable route. It is also the slowest and the most expensive per design. Here is what software can actually do instead, and where it stops.

Ruby Kinglet makes one of the tools below and we say plainly which jobs still need a person. Updated July 2026.

Why this is hard

A sketch or a render is a picture. It shows one view, it has no dimensions, and it does not know what is behind the stone. A castable file needs things the picture never contained: real wall thickness, prong dimensions that survive setting and wear, correct stone seats and undercuts, a shank in a real ring size, and a hollow back that does not trap investment. Software has to invent all of that. Some of it can be inferred sensibly. Some of it is a guess, and a guess that looks fine on screen fails at the caster.

The three routes

A freelance CAD jeweler takes your sketch and models it. Typical turnaround is a few days, and you get real parametric geometry plus a person who will tell you when the design cannot be made. This is the default for a reason. A generic image-to-3D tool, the kind not built for jewelry, gives you a mesh in seconds that is almost never castable: surfaces are approximate, thicknesses are decorative, and stone seats do not exist. A jewelry-specific AI tool sits in between, and that is the interesting part of the market.

What each route produces

What the AI route does well today

Photo or sketch to a printable mesh is production-grade for most standard rings, pendants, and studs. A file in under a minute, castable after a thickness check on a lot of pieces. That is a real change from three years ago and it genuinely removes a step for simple work. Where it still falls down is exactly where you would expect: micro-pave, unusual settings, anything with a lot of small parts holding stones. On those, a mesh that looks right on screen will hand your caster a problem, and a CAD jeweler is not optional.

Mesh or parametric, and why it matters here

An STL is a frozen mesh. It prints. It does not resize, and you cannot swap a head or change a shank without rebuilding. If the piece is one-off, that is fine. If the customer wants it in a size 7 instead of a 6, or in yellow instead of white, or with a 1.5 carat instead of a 1, you are back at the start. A 3DM is parametric, so those are edits rather than rebuilds. Anyone selling variations of the same design should care about this a lot more than the price difference suggests.

How Ruby Kinglet handles it

Two separate jobs, and they are worth keeping separate. Photo to STL produces a printable, manufacturable mesh with a thickness check, plus castable component parts you can compose, and it is on the $35 Professional plan. The 3DM side is on the $179 Studio plan, or $125 a month billed annually, and it does two things: edit an existing 3DM into a new piece, so a new design starts from past work rather than from scratch, and render one 3DM many ways with swappable metals, stones, and settings while the geometry stays intact. Both run on a Rhino parametric engine with jewelry-specific libraries. The honest limit: that engine is newly released, so it does its best work on common styles like engagement rings and is still experimental across the broader range.

In the app these are separate named workflows rather than one button, more than a dozen of them once you count the ones outside the CAD tab: single photo to .STL, multi-photo for more precision, editing an existing .STL, repairing and prepping an .STL for casting, editing a .3DM into variations, parametric .STL sized in millimetres, swappable-part renders, seeding a photoshoot from a CAD file, capturing a reference angle off a posed CAD model, and rendering an exact studio photo per storefront combination straight from a .3dm. Several are written up step by step in our documented generative workflows, with more coming as the engine grows, including composing cast .STL parts into a .3DM. The repair-and-prep flow is the one most relevant here: it takes a mesh that looks right and gets it to a state a caster will accept. And the same caveat we would give about any tool, ours included: test the file against your own manufacturing before you rely on it. A render that looks right is not the same as a file that casts clean.

Choose by fit

For the wider mesh-versus-parametric picture, see what is actually manufacturable in 2026. For one complex piece where being wrong is expensive, hire a CAD jeweler. For a printable mesh from a sketch, fast and cheap, Pencil or JewelArchitect. For volume on common styles, with editable geometry and variations, a jewelry-specific tool with a parametric engine. For full control across every category, learn Rhino or MatrixGold. Do not use a general image-to-3D tool for jewelry you intend to cast.