Skip to content

Global-AI-Group-Inc/content-generation-skills / skills/mira-3d

mira-3d

Verified

3D models on Mira AI: a textured mesh (GLB, FBX, OBJ, USDZ) from a text prompt or from one to four photos, optionally rigged as a humanoid with animation clips. Use it for "3D model", "3д модель", "сделай меш", "GLB", "для Blender", "для Unity", "for a game", "3D print", "AR product", "turn this photo into 3D", "rig this character". Pass skill="3d" to generate_3d, or call rig_3d on an existing model. Do NOT use for 2D renders that merely look three-dimensional (mira-toon3d) or for product photos (mira-product).

Mira 3D

One subject becomes one mesh. The model reads the prompt or the photo as a description of a single object seen from all sides, so everything that helps it understand the shape belongs in the request, and everything that describes a scene, a mood or a camera does not.

When to use

A character, a prop, a product or a piece of environment that has to exist as geometry: to drop into Blender, Unity, Unreal or a browser game, to print, to show in AR, to animate as a rigged humanoid. Anything where the deliverable is a file with polygons, not a picture.

Doctrine

This is a single-asset 3D generation: one subject, complete on all sides, standing alone on a neutral background with nothing else in frame. Describe the object as a physical thing: its overall silhouette, the parts it is made of, their proportions to each other, the materials and colours of each part, and the surface finish. Name the category first (a humanoid character, a creature, a vehicle, a prop, a piece of furniture, a building fragment) so the model knows what kind of shape to expect. For characters and creatures, ask for a symmetrical A-pose or T-pose with arms away from the body and legs apart, limbs fully visible, no held props overlapping the body, so the mesh can be rigged. For products and props, give the real-world size or the scale relative to a hand, and keep the label or logo geometry simple. Say nothing about lighting, camera angle, depth of field, motion or atmosphere: none of it exists in a mesh. Do not ask for text, lettering or fine print on the surface. Write the prompt in English.

Interview

  • subject What kind of asset is it? (default: prop)
    • character Character: A humanoid or creature, symmetrical A-pose, arms away from the body, legs apart, limbs fully visible, no props overlapping the body.
    • prop Prop or product: A single object with a clear silhouette, real-world proportions, standing on its base, complete on all sides.
    • environment Environment piece: A modular piece of a set, such as a wall, a tree, a rock, a vehicle or a building fragment, with a flat resting side so it can be placed.
  • use Where will it go? (default: general)
    • general General: Clean geometry and a full-colour texture, the default for previews and renders.
    • game Game engine: Clean quad-friendly topology at a modest polygon count, no floating pieces, a texture that reads at a distance.
    • print 3D print: A single watertight solid without thin protrusions, colour is irrelevant.
  • detail How much detail? (default: standard)
    • draft Draft: A quick low-detail pass to check the shape.
    • standard Standard: Full geometry and texture at the model's normal quality.
    • high High: The model's highest geometry and texture setting, with PBR maps where the model offers them.

Hint

Name the object, its parts and materials, and say if it is a character that should be rigged.

Model picks

hunyuan-3.1 is the default: strong geometry from text or a photo, FBX and USDZ included, PBR on request. meshy-7 for characters that must be rigged and animated, and for game-ready low-poly meshes. trellis-2 for props and products from a photo with the cleanest UVs. rodin-fast for a cheap draft in any format, including STL for printing. Rigging works on any finished GLB through rig_3d, whichever model made it, and on a mesh the user already has: from Blender or another DCC tool, export the character as GLB, create_upload it (kind model3d) and call rig_3d with the returned libraryItemId.

Checklist

  • One subject, nothing else in the prompt or the photo.
  • Characters in A/T-pose with limbs apart when a rig is planned.
  • Materials and colours named per part, not as a mood.
  • Photo references show the object cleanly, ideally on a plain background.

Do not

Do not describe a scene, a camera or a light. Do not ask for text on the surface. Do not send a photo of several objects and expect one of them back.

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall