Zetta N1 — every way into 3D, on every plan
Zetta N1
Our original neural model. Text, a single image, or up to seven reference views become a mesh — plus low-poly, sketch and untextured modes that no other model here offers.
What it does
Every input shape
A bare prompt with no image, a single photograph, or multiple reference views. N1 is the only model that accepts all three.
Multi-view reconstruction
Up to seven named view slots — back, left, right, top, bottom and two three-quarter angles — so the far side of a model is observed rather than inferred.
Low-poly and sketch modes
Generate game-ready low-polygon geometry directly, or feed a line drawing and have it read as a drawing rather than as a photograph.
Geometry-only output
An untextured white model, when you intend to texture it yourself and do not want to pay for a texture you will throw away.
How it works
- Choose your inputA prompt, an image, or a set of views. A prompt is converted to a reference image first, which is why prompt wording that changes the picture changes the mesh.
- Pick a base and modifiersTextured, low-poly or geometry-only, with PBR materials and a target polygon count where the base supports them.
- GenerateThe job is priced and quoted before it runs, including any format conversion it needs.
- RefineRetopologise, unwrap, texture, segment or rig the result without leaving the studio.
What people use it for
- Anything where you have several reference angles and want the back to be right
- Game-ready low-poly assets generated directly rather than reduced afterwards
- Concept exploration from a written description, before any art exists
- Sketch-to-3D from concept drawings and line art
- Untextured base meshes destined for hand-authored materials
What it cannot do
- Text and lettering on a surface come back as letter-shaped marks rather than the letters you asked for
- Thin repeated structures — chain links, railings, spokes — reconstruct fused
- Interiors are inferred, so mugs come back solid and boxes do not open
- Nothing measures anything: exact dimensions are not achievable
- Transparent and reflective objects reconstruct badly, because the camera recorded the reflection rather than the surface
The details
- Inputs
- Prompt, single image, or multi-view
- View slots
- Up to 7, plus the primary image
- Bases
- Textured, low-poly, geometry-only
- Modes
- PBR materials, custom face count, sketch
- Availability
- Every plan, including free
- Output
- GLB, with OBJ and FBX available