Tripo Segmentation: The Essentials

Tripo Segmentation takes a finished 3D mesh and splits it into separate, editable parts, so a single fused model becomes a set of components you can select, texture, rig, or swap on their own. This is the step between generating a 3D asset and actually working with it in a production pipeline.

The family has two members. Tripo Segmentation v1 is the geometry-based version documented here: it analyzes a model's topology to find natural separation points and divides it into parts automatically, with no settings to configure. Tripo Segmentation v2 is a newer semantic version (currently in preview) that adds controllable granularity and optional reference-image guidance. Both versions are covered below.

Which Model Should I Use?

Reach for v1 when you want a quick, hands-off split and the model has reasonably distinct parts. Choose v2 when you need to control how finely the mesh is divided or guide the split with a reference image.

Parameters

Tripo Segmentation v1

v1 is fully automatic. It exposes a single input and no tuning parameters.

There is no granularity or connectivity control in v1: those live in v2. If you need to decide how many parts the model is broken into, use v2.

Tripo Segmentation v2 (preview)

v2 adds semantic segmentation plus controls for how the mesh is divided.

How It Works

Upload a 3D mesh and Tripo Segmentation v1 analyzes its geometry, detecting where one part naturally ends and another begins. It returns the same model with its parts separated into individual components, ready to be selected and edited on their own. The result keeps clean, organized geometry, so the parts stay compatible with rigging, texturing, and export.

The segmentation is structural: the parts are split in the model data, but a normal render still shows the original textured model. To see and select the individual parts, open the segmented asset in the Scenario 3D viewer. The examples below were produced by feeding meshes from Tripo P1, Tripo 3.1, and Hunyuan 3D 3.1 Pro into Tripo Segmentation v1.

A robot crab (generated with Tripo P1), segmented by v1. Open on Scenario to rotate the mesh and inspect the separated parts.

Robot crab · Open on Scenario

A swordsman statue (generated with Tripo 3.1), segmented into figure and sword and armor pieces.

Swordsman statue · Open on Scenario

A table scene with props (generated with Hunyuan 3D 3.1 Pro), where each object becomes its own part.

Table with props · Open on Scenario

Semantic Segmentation and Granularity (v2)

Tripo Segmentation v2 replaces v1's pure geometry analysis with semantic understanding, so the parts line up with meaningful components (limbs, accessories, mechanical pieces) and the boundaries come out sharper. It also gives you controls that v1 does not.

Granularity sets how finely the mesh is divided: Simple yields a few large parts, Balanced (the default) sits in the middle, and Detailed breaks the model into many smaller parts. Split By Connectivity additionally separates pieces that are not physically joined. A Reference Image can guide the split and overrides the Granularity setting when supplied.

As with v1, the split is structural, so a flat render looks like the original model. Open each example on Scenario to rotate it and inspect the parts produced at each level.

Robot crab, v2 at Detailed granularity · Open on Scenario

See each granularity level on Scenario: Simple (chair), Balanced (fox), and Detailed (robot crab).

Use Cases

• Rigging prep: separate limbs, joints, and accessories so a character mesh is ready to weight and rig.

• Part-based texturing: isolate a component to retexture or recolor it without touching the rest of the model.

• Model customization and swapping: break a character or prop into parts so pieces can be replaced, hidden, or mixed and matched.

• Hard-surface and mechanical assets: split robots, vehicles, and machines into their mechanical components for editing or animation.

• 3D printing: divide a model into printable pieces that can be produced separately and assembled.

• Modular asset pipelines: turn a single generated mesh into a kit of reusable parts.

Tips for Better Results

1. Start from a clean mesh. Models with well-defined, distinct parts (clear limbs, separate accessories) segment more cleanly than smooth, fused shapes.

2. Keep files under 150 MB. That is the hard input limit; decimate or retopologize very dense meshes first if needed.

3. Inspect parts in the Scenario viewer. A flat render looks unchanged, so open the segmented asset to confirm the parts split where you expect.

4. Use v1 for speed, v2 for control. If v1's automatic split is too coarse or too fine for your model, switch to v2 and set the granularity.

5. Segment before rigging or texturing. Doing the split first gives you the separate components those later steps depend on.

Known Limitations

• No granularity control. v1 decides the number of parts automatically. For Simple, Balanced, or Detailed control, use v2.

• Geometry-based splits can be uneven. On complex or organic models the part boundaries can be fuzzy or inconsistent. v2's semantic approach produces sharper boundaries.

• Parts are not color-coded in a standard render. The segmentation is structural, so a normal preview looks like the original model. Open the asset in the Scenario 3D viewer to see and select the parts.

• Input only, no reference image. v1 takes just the mesh. Reference-image guidance is a v2 feature.

• Scenario exposes only the mesh input. The provider tool accepts OBJ, FBX, and GLB; on Scenario the model takes the single 3D file input documented above.