Turning a concept into a detailed 3D body

Some things are hard to sculpt by hand: a cartoon dinosaur with a smile and toes, a dragon with folded wings, a little statue of your dog. For these, Solid Vibes can generate the body from pictures — your concept sheet, a few photos of the real thing, or a picture it draws first from your description. The result looks like the pictures (faces, toes, spikes, folds), and it is a normal solid in your recipe: you can resize it with sliders, cut holes and pockets into it, hollow it, add a flat base, bosses or text, and print it.

Two ways in

1. The agent picks it for you

When you build a character, figurine, creature, animal or art object from a concept sheet (Concept → Build in 3D) or from a reference picture you attach, the agent generates the body instead of sculpting it, then adds the rest as normal: a flat base so it stands, and whatever you asked for — a magnet pocket, a battery box, a hole for a keyring, your name on the base.

A happy toy dinosaur for my 4 year old, 180 mm long, made of PLA, with a magnet pocket under each foot.

You can also ask for it mid-conversation, with a photo or just words:

Add a little dragon sitting on top of this box, about 6 cm tall.

The agent never uses it for functional or precise parts (brackets, boxes, mounts, gears) — those stay sketches, extrusions and holes — and for simple smooth forms (a pebble, a mouse, a vase) it keeps sculpting, because a sculpted shape stays editable with sliders and edits.

In the agent trace you'll see Generated the shape from your concept sheet (or your photos / your description) with the number of triangles, the size and how well its outline matches the concept.

2. Generate it yourself (⌘K)

Press ⌘K and choose Generate a 3D model… (or Model this from a photo…). A small window opens:

• Photos: drop or pick 1–4 photos of one object — front, side and three-quarter views on a plain background work best. Pictures already attached to the prompt can be used too.
• Or a description: "a small friendly dragon sitting with its wings folded, about 8 cm tall". Solid Vibes first draws a picture of it (a normal concept image), then turns that picture into 3D.

Press Generate. It takes about a minute or two; the body arrives as a preview you Accept like any agent change. It goes into the Thing you're in — the empty piece if there is one, otherwise a new piece next to your model. From Home → Create, Generate a 3D model makes a new Thing and opens the same window.

What you can still change

The generated body is a mesh body, stored in the recipe as Generated body from …. On it you can:

• Resize it with its length / width / height sliders (one or all of them; the body stretches to fit).
• Flat base: cut the bottom flat so it stands on the print bed (the agent does this for you; the Add a flat base chip shows when it's missing).
• Holes and pockets: a magnet pocket, a battery box, a keyring hole, a slot for a stand — any sketch + cut or a box / cylinder cut.
• Add parts: bosses, a peg, a base plate, a name plate, raised or engraved text on a flat face.
• Reshape it with shape sliders: "make the neck longer", "a bigger head" — the agent (or Mesh tools → Shape slider) puts a named slider around that part, and you drag it in Parameters.
• Smooth it, make it symmetric, thicken thin bits with Mesh tools (⌘K → Mesh tools…, right-click the body, or the body's ⋯ in the Recipe).
• Hollow it for printing: Mesh tools → Hollow empties the inside with walls of the thickness you choose and drills drain holes at the lowest points (resin needs them).
• Split it into parts that fit your printer, with keyed pins, each part laid flat on its cut.
• Pockets for magnets and threaded inserts at a spot you click.

See Working with meshes for every tool. Sculpt mode doesn't apply (there's no sculpting cage): it opens Mesh tools instead. The smoothness check reports it as smooth as generated (the edges between its triangles aren't creases to fix), or bumpy in places with a pointer to Smooth.

Printing tips

• The agent adds a flat base about 1–2 mm deep; keep it, or the figure will rock on the bed.
• Thin, delicate details (whiskers, spikes, fingers) print best in resin or at a small nozzle; with FDM, ask for a bigger size or "chunky, toy-like details" in the description.
• Hollowing saves a lot of material on big figures: 2 mm walls work well for FDM and resin. Mesh tools → Hollow adds the drain hole at the lowest point for you (two for a long body).
• Bigger than your printer? Mesh tools → Split suggests where to cut and adds pins so the parts line up; each part prints flat on its cut.
• Overhangs under the chin, belly or tail may need supports; the printability check in Make shows where.

Credits

A generated body costs 150 credits (Fast or Smart) and is charged only when a body is delivered — if the generation fails or the pictures can't be used, nothing is charged. With a description instead of photos, the picture it draws first is a normal concept image (5 in Fast, 8 in Smart). Inside a build, the agent's own work (the flat base, holes, the look check) is charged as a normal build.

In the free trial, your first generated body is free. The trial's 100 credits are fewer than a body's 150, so the first body you get during the trial uses no credits at all: the Generate a 3D model… card says First one free during your trial, and the agent tells you "Your first generated body is on us during the trial." when it uses it. It's used up only when a body is delivered (a failed generation keeps it), and it's one per account. A build around it (the flat base, holes) and a picture drawn from a description are charged as usual. After that, the normal 150 credits apply: with fewer, the agent shapes organic forms with its sculpt tools instead and says why, and Generate a 3D model… tells you how many credits it needs.

The credits are checked before anything starts. Without enough of them you get the usual credits prompt; on Free (no monthly credits once the trial's are used up) both Generate a 3D model… and the agent show the AI needs a subscription dialog with the plans instead of starting. See Credits, explained and Plans and limits.

How it works

The pictures go to an image-to-3D model on fal.ai (US): Meshy 7.1 (several views, the default) or Microsoft TRELLIS.2 (one view) — your workspace admin chooses. The result is turned into one closed, printable body of about 20,000 triangles, stood upright and turned to face the way your concept sheet shows, scaled to the size you gave (limits like a build height only cap it), and placed on the bed. With a concept sheet, its outline is compared with the sheet's front, side and top views, like the agent's own look check.

Limits

• One object per generation. Several objects, busy backgrounds or a hand holding the thing confuse it.
• Colours and textures aren't kept: the body is geometry only (you pick material and colour in Make).
• Very thin parts can merge or disappear when the body is made printable (anything thinner than about half a millimetre at 18 cm long).
• There's no sculpting cage: reshape it with shape sliders (stretch, grow, move or bend a part), smoothing and make symmetric — not by pulling individual points of the surface.
• It needs your workspace's AI features to include the generator; if you don't see Generate a 3D model… in ⌘K, it isn't available in your deployment.