Skip to main content

Scene Insight

Scene Insight is an Unreal Engine editor tool that visualizes and analyzes level performance by spatial zones (clusters). It bakes per-zone metrics — draw calls, triangles, materials, lights, texture memory and dozens more — paints a color-coded heatmap directly in the viewport, and lists every zone in a sortable table you can filter, snapshot, compare and export.

Scene Insight viewport heatmap — color-coded cluster boxes over a level

At a glance
  • Engine: Unreal Engine 5.4+
  • Category: Performance
  • Type: Editor tooling (static / bake-time analysis — no runtime cost in your shipped game)
  • Modules: SceneInsightCore (runtime types), SceneInsightEditor (editor UI)

Why zones instead of a single number?​

Profiling tools usually give you a single per-frame figure for the whole view. That tells you that a level is heavy, but not where. Scene Insight partitions your level into spatial clusters and scores each one against a performance budget, so you can answer questions like:

  • Which corner of the map blows the draw-call budget?
  • Where are 4K textures or shadow-casting lights piling up?
  • Which zone regressed between this week's build and last week's?

What it measures​

Scene Insight collects a deep set of static metrics per cluster, grouped into tiers:

GroupExamples
GeometryTriangles, vertices, static/skeletal mesh counts, Nanite ratio, LOD-saving potential, duplicate-mesh / ISM-conversion potential
MaterialsUnique materials, material slots, opaque/masked/translucent breakdown, master-to-instance ratio, custom HLSL, two-sided, subsurface
LightingStatic/stationary/movable counts, point/spot/rect/directional, shadow casters, overlapping-lights heuristic, IES & light functions, lightmap resolution
MemoryTexture memory, lightmap memory, mesh memory, audio memory, max texture dimension, virtual textures
FX / MiscNiagara CPU/GPU emitters, Cascade systems, decals, foliage instances, physics actors & constraints, ticking actors

Every cluster gets a single composite score computed against the active budget profile: 0.0 = healthy (green), 1.0 = exactly at budget, >1.5 = critically over budget (red).

The workflow​

Choose generators → Bake → Read the heatmap → Drill into the table
        → Snapshot a baseline → Compare later → Export a report
  1. Generators decide how the level is partitioned into clusters (grid, octree, World Partition cells, tags/folders, or hand-placed volumes).
  2. Bake runs the generators and the static collector to fill in every metric.
  3. The viewport heatmap and cluster table show the results.
  4. Snapshots let you save a baseline and diff future bakes against it.
  5. Export produces CSV / JSON / Markdown / HTML reports for code review or CI.

Next steps​