A well-structured prompt is the fastest way to get a usable result out of Meshy. This guide breaks down the anatomy of a good prompt and gives templates for the most common asset types.

A strong prompt generally covers: the subject, its material or surface (e.g. "worn leather," "brushed steel"), its style (realistic, stylized, low-poly), and any structural details you care about (pose, proportions, symmetry).

"A [material] [object], [style], [condition/detail]." Example: "A weathered wooden treasure chest, stylized game asset, with brass hinges and a rounded lid."

"A [descriptor] [character type], wearing [outfit], [pose], [style]." Example: "A stocky dwarf blacksmith, wearing a leather apron, standing pose, stylized fantasy game art."

"A [era/style] [vehicle type], [material/finish], [condition]." Example: "A retro-futuristic hover car, chrome and glass finish, clean showroom condition."

"A [building type], [architectural style], [material], [scale note]." Example: "A small stone cottage, medieval European style, thatched roof, single-story scale."

For print-ready assets, add print-friendly cues to your prompt, such as "solid base," "no thin overhangs," or "single connected piece," to reduce cleanup before slicing.

Do keep prompts specific but concise. Don't stack too many conflicting style words (e.g. "photorealistic cartoon") since that tends to produce muddled results.

Use the negative prompt field to explicitly exclude unwanted elements, such as extra limbs, floating geometry, or a specific material you don't want.

For characters, you can request specific poses (idle, walking, action) directly in the prompt to get a result closer to your intended use without manual rigging first.

A sentence or two is usually enough, covering subject, material, and style. Overly long prompts with conflicting details can confuse the result.

You can, but conflicting styles (like "photorealistic" and "cartoon") tend to produce inconsistent results. Pick one dominant style per prompt.

These templates are aimed at text-to-3D. Image-to-3D relies primarily on your reference image, though a short prompt can still help clarify style or material.