DecimateMesh - ComfyUI Built-in Node Documentation

DecimateMesh simplifies a 3D mesh to a target face count using quadric error metric (QEM) simplification, running the computation on the active compute device. The "midpoint" placement mode is the cumesh-faithful preset providing the best quality while preserving thin features such as hair, whereas "qem" places vertices at the QEM-optimal position with optional line and feature-edge controls. The output mesh stays welded.

Inputs

Common Inputs

Parameter	Description	Data Type	Required	Range
`mesh`	The 3D mesh to simplify.	MESH	Yes	-
`target_face_count`	Target max faces. 0 disables. (default: 200000)	INT	Yes	0 to 50000000
`placement_mode`	midpoint: cumesh-faithful (recommended). qem: QEM-optimal placement. (default: `"midpoint"`)	DYNAMIC_COMBO	Yes	`"midpoint"` `"qem"`

Midpoint Inputs

The "midpoint" placement mode does not expose additional sub-parameters; it uses the default midpoint placement preset.

QEM Inputs

The following sub-parameters appear in the interface only when placement_mode is set to "qem".

Parameter	Description	Data Type	Required	Range
`line_quadric_weight`	Per-edge line-quadric weight; preserves sharp ridges/valleys. 0 = off. (default: 0.0)	FLOAT	No	0.0 to 100.0
`feature_edge_quadric_weight`	Extra quadric weight on dihedral feature edges (creases). 0 = off. (default: 0.0)	FLOAT	No	0.0 to 1000.0
`feature_edge_min_dihedral_deg`	Minimum dihedral angle (degree) to count an edge as a feature edge. (default: 30.0)	FLOAT	No	0.0 to 180.0
`clamp_v_to_edge`	Project the QEM-optimal position onto the collapsed edge segment. (default: true)	BOOLEAN	No	`true` `false`

Note: Decimation is skipped when target_face_count is 0 or when the mesh already has fewer faces than the target. The node displays a face reduction summary on itself, for example faces: 1.23M → 200K (-84%).

Outputs

Output Name	Description	Data Type
`mesh`	The simplified mesh with the reduced face count; connectivity stays welded.	MESH

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): 55336e5b52e27d940e5402ecd74fd0ac847a1c6acd35955eccf72aab8ed940f9

​Inputs

​Common Inputs

​Midpoint Inputs

​QEM Inputs

​Outputs

Inputs

Common Inputs

Midpoint Inputs

QEM Inputs

Outputs