LOD Batch Studio

Build LOD copies. Measure the result.

Preflight selected static meshes, generate isolated decimated LOD copies at three ratios, measure actual triangles, retain source meshes, isolate levels and export budget reports.

• Three native decimation ratios
• Actual triangle counts
• Independent source-preserving copies
• Level visibility preview

Preflight selected static meshes, generate isolated decimated LOD copies at three ratios, measure actual triangles, retain source meshes, isolate levels and export budget reports.

Windows x64. Intended Blender builds: 4.5.15 LTS and 5.2.3 LTS. See the included verification notes for completed verification; version targeting alone is not evidence of a passed test.
In Blender choose Edit > Preferences > Get Extensions > menu > Install from Disk. Select lod_batch_studio-1.0.0.zip without extracting it, and enable the extension.
In the 3D Viewport press N and open the LOD Batch tab. No account, activation, external Python, subscription or network service is required.
Save a backup before bulk operations. Example opens the bundled scene only after you have saved your current changes. Guide opens this local offline documentation in your browser.

1. Select 1–100 unparented static local mesh objects. Apply modifiers on a reviewed copy first. Animation, rigs, shape keys, constraints, parenting, instancing and object-linked material slots are blocked. Each source needs at least 12 triangles and at most 250,000 polygons / 500,000 vertices.
2. Enter three strictly decreasing ratios between zero and one, for example 0.5 / 0.25 / 0.1. Enter a triangle warning budget. Ratios request decimation; final triangle counts are measured, not promised. A batch is capped at 500,000 source triangles. Generation runs synchronously in Blender; use smaller batches to reduce waiting.
3. Analyze checks eligibility and shows source counts. Generate creates a unique LOD_Review collection containing three aligned independent mesh objects per source, simplified with Blender's native Collapse decimation and triangulation. Source geometry, transforms and visibility remain unchanged.
4. All LOD copies initially overlap at the source transform. Hide the source manually for inspection. Set Preview level to 1, 2 or 3 and click Preview chosen level to isolate generated levels. Level 0 shows all generated levels. Preview changes generated-object viewport and render visibility; it does not change source visibility. Restore batch visibility before switching preview levels.
5. Review actual triangle counts and over-budget flags in the report. Inspect silhouettes, UVs, normals and shading. Decimation can create visible artifacts and does not promise retopology, preserved UV seams, watertightness or target-engine LOD switching. No FBX/GLB export or engine setup is performed by this plugin.
6. Generated collections are retained so you can inspect and edit them. Blender Undo can undo generation. There is no destructive automatic batch-delete button. A new Generate action creates another uniquely named collection. Preview references are guarded against renamed or removed objects.
7. Export the offline audit with exact counts for all levels. This is a static LOD preparation tool, not runtime frame-rate certification.

Choose an existing output folder and click Export offline report after analysis or completion. Every export uses a new timestamp/UUID folder. HTML and JSON are offline, with escaped user data in the HTML. Reports do not overwrite your existing files. They may contain private local paths and object names; review before sharing.
Settings, reports and supported restoration records are saved with your .blend file. Blender Undo applies to scene-data operations, not files written to disk. Keep separate project backups. None of the tools runs a background network service or deletes your source files.

GPL-3.0-or-later, complete original Python source included. Original example scene included. Blender itself is not included. Use Blendkit creator messaging for support; provide your Blender version, exact error and a minimal non-confidential reproduction project.
No income, runtime performance, target-engine compatibility or arbitrary-scene conversion guarantee is made. Review actual results before delivery.