SuperAPI AI Directory

lod-pipeline

Design & creation

View repository

Updated 10-31 12:24

About this skill

LOD Pipeline

Intentional LOD chains. Preserve silhouette. Validate reductions.

When to Use

  • Game assets needing LOD0–N
  • Batch LOD for modular kits
  • Before Unity/Unreal/Godot export

Workflow

Approve LOD0 → Generate LOD1..N → Fix silhouette breaks
  → Naming → Screen-size table → Export verify

Typical Budgets (guide)

LOD Tris vs LOD0 Keep
LOD0 100% Full detail
LOD1 40–60% Major forms
LOD2 15–30% Silhouette
LOD3 5–10% Blob / impostor candidate

Tune per asset importance and camera.

Generation Methods

  1. Decimate (planar/collapse) — fast props
  2. Manual retopo low — hero characters/vehicles
  3. Delete interiors / small parts — vehicles, buildings
  4. Bake normals from LOD0 → low — when detail must move to texture

Naming

SM_Asset_LOD0
SM_Asset_LOD1
SM_Asset_LOD2

Collection: COL_LODs_[Asset]

Validation Checklist

  • Silhouette matches at intended distance
  • No major holes / inverted normals
  • UVs still valid (or rebaked)
  • Materials slots consistent across LODs
  • Vertex colors/UVs needed by shader preserved
  • LOD transitions planned (screen size %)

Screen Size Starters

LOD Screen size (UE-like guide)
LOD0 1.0 – 0.5
LOD1 0.5 – 0.25
LOD2 0.25 – 0.1
LOD3 0.1 – 0.0

Constraints

MUST DO

  • Protect silhouette before poly reduction
  • Keep LOD0 as source of truth
  • Document budgets in production brief

MUST NOT DO

  • Write code or snippets for Blender versions before 5.2, or add version fallbacks: this plugin targets 5.2+, and legacy branches teach users APIs that no longer exist
  • Blind decimate until unrecognizable
  • Different material slot counts across LODs without reason

Scripts

Tested on Blender 5.2 LTS via the official Blender MCP. Run with execute_blender_code (or headless, see the blender-mcp-workflow skill); read the script header for every argument. When the user will paste code into Blender themselves, give them the short runpy call with the absolute path of this installed script and their ARGS. Don't inline or rewrite the script: the tested file is the point.

Script What it does
scripts/generate_lods.py Generate LOD copies of a mesh with Decimate (collapse) at given ratios. The source becomes/keeps LOD0; copies are named _LOD1.. and placed in a COL_LODs collection. Copies share nothing with the source (single-user data).
import runpy
result = runpy.run_path("<skill_dir>/scripts/generate_lods.py", init_globals={"ARGS": {...}})["result"]

ARGS: object, ratios, apply, rename_source, offset_x

Reference Guide

Topic Reference Load When
Lod Budgets references/lod-budgets.md When needed
Lod Validation references/lod-validation.md When needed