LOD (Level of Detail) in Unity3D

LOD stands for Level of Detail. In Unity, LOD systems let you swap between different versions of a model depending on how large it appears on screen.

The basic idea is:

  • Close objects use a high-detail mesh.
  • Medium-distance objects use a simpler mesh.
  • Far-away objects use an even simpler version.

This helps save rendering cost without making nearby objects look worse.

Recommended LOD Workflow

Use LODs on detailed objects that are visible from multiple distances, especially repeated environment assets.

  1. Pick one repeated or large detailed asset that stays visible at distance.
  2. Create cheaper mesh versions while preserving the silhouette and major forms.
  3. Add a Unity LOD Group, test transitions in the real scene, then repeat only where the benefit is clear.
VRChat note

For VRChat worlds, LODs are most useful on environment assets players can see from different distances. Test with realistic player viewpoints, avatar load, mirrors, and target platforms before assuming the setup helped.

Why LODs Matter

If the player is far away from an object, they usually cannot see small mesh details clearly. Keeping the full high-detail version active at all distances often wastes performance.

LODs help by:

  • Reducing geometry cost at distance.
  • Making larger scenes more manageable.
  • Keeping nearby objects detailed where it actually matters.
  • Letting repeated props and architecture become cheaper when they are no longer the focus.

LOD work is about spending detail where players can actually see it. If an object is always close, always hidden, or already very simple, an LOD setup may add complexity without much benefit.

LODs vs Occlusion Culling

LODs and occlusion culling solve different problems:

Technique What it does Best use
LOD Swaps an object to cheaper versions as it becomes smaller on screen Objects still visible, but farther away.
Occlusion culling Stops rendering objects blocked by walls or other occluders Objects not visible from the current viewpoint.
Manual scene layout Uses rooms, corners, terrain, or walls to limit what can be seen Large worlds where not everything should be visible at once.

Use them together when it makes sense. An object might be hidden by occlusion in one area, then use LODs when it becomes visible at a long distance.

What an LOD Group Does

In Unity, an LOD Group controls which version of a model is shown at different screen sizes or distances.

A typical setup might include:

  • LOD0 for the closest and highest-detail mesh
  • LOD1 for a medium-detail version
  • LOD2 for a lower-detail version

Some setups may also end with a very low-detail version or eventually cull the object at extreme distance depending on the use case.

What Each LOD Should Preserve

Level What it should preserve What can usually be simplified
LOD0 Full close-up shape, important bevels, readable detail Nothing important; this is the best version.
LOD1 Main silhouette, major surface breaks, recognizable shape Small bevels, dense minor loops, tiny inset details.
LOD2 Broad silhouette and color/material readability Most small geometry, tiny details, hidden back-side complexity.
Culled Nothing rendered Use only when the object no longer matters at that distance.

The lower LODs do not need to look perfect up close. They need to look believable at the distance where Unity shows them.

When LODs Are Worth Using

LODs are most useful for:

  • Larger props.
  • Repeated environment objects.
  • Detailed architecture.
  • Objects seen from many distances.
  • Trees, rocks, modular building pieces, distant decorative structures, and background set dressing.

They are usually less important for:

  • Very small objects.
  • Already simple models.
  • Assets only seen at close range.
  • Objects hidden inside rooms where occlusion or layout already handles visibility.

Before creating LODs, ask whether the asset is expensive enough and visible enough to justify the extra setup. Repeated assets usually give the best return because one good LOD setup can help many copies.

A Good Beginner Workflow

The usual process looks like this:

  1. Prepare the different mesh detail levels.
  2. Name them clearly, such as Rock_A_LOD0, Rock_A_LOD1, and Rock_A_LOD2.
  3. Import them into Unity.
  4. Add an LOD Group to the object.
  5. Assign the meshes to the different LOD slots.
  6. Test the transitions in the scene.
  7. Adjust transition distances only after viewing the object from normal player positions.

The LOD Group is where Unity decides when to switch from one version to another.

How to Think About LOD Quality

Beginners sometimes assume every LOD needs to look almost identical up close. That is not the goal.

The real goal is:

  • LOD0 looks good near the camera
  • lower LODs look acceptable at the distance they are used
  • transitions do not feel distracting

If a lower-detail mesh looks slightly worse up close, that is often fine, because the player should not normally be seeing that version at close range.

Setting Up LODs Carefully

When assigning LOD levels, think about:

  • The overall silhouette.
  • Major readable forms.
  • Whether the switch feels obvious.
  • Whether materials and UVs still look acceptable.
  • Whether the object is important enough to justify multiple LODs.

This is not just about adding more steps for the sake of it. It is about choosing the right tradeoff between quality and performance.

VRChat LOD Testing Checklist

After setting up an LOD Group, check:

  • The object still looks good from normal VRChat player height.
  • The LOD switch is not distracting during normal movement.
  • The lower LOD keeps the broad silhouette readable.
  • Repeated copies actually benefit from the setup.
  • The object behaves acceptably near mirrors and busy social areas.
  • Quest or Android targets still get an appropriate version of the asset.
  • The LOD meshes are named clearly enough to maintain later.

Do not only inspect the object in a blank test scene. Put it where players will actually see it.

Test the LOD Transitions

After setting up the LOD Group:

  1. Move the camera toward and away from the object
  2. Watch for visible popping
  3. Check whether the silhouette collapses too much
  4. Adjust the transition points if needed

If the switch is too noticeable, it may mean:

  • the lower LOD is too simplified
  • the transition happens too early
  • the differences between levels are too abrupt

Common Beginner Mistakes

Using LODs on the wrong objects.

Not every asset needs an LOD setup. Very small, already simple, or close-only objects may not benefit enough to justify extra meshes and setup time.

Making lower-detail versions too aggressive.

This can make the object visibly collapse or flicker in quality when switching. Preserve the silhouette first, then simplify smaller details.

Not testing transitions in the real scene.

An LOD may look fine in isolation but feel wrong once placed in the environment. Test while walking through the actual room, path, or outdoor area.

Forgetting that repeated objects benefit most.

A single detailed object may matter less than a repeated prop used dozens of times. Prioritize repeated trees, rocks, modular pieces, or architectural assets first.

Practical Advice

  • Use LODs where the object is detailed and seen from multiple distances
  • Focus on repeated or large assets first.
  • Preserve silhouette before small surface detail.
  • Test in the real scene, not only in an isolated model preview.
  • Keep naming and mesh setup clean so the LOD workflow stays manageable.
  • Compare LOD work against other optimization tools such as occlusion culling, material cleanup, and texture reduction.

Relationship to Performance Work

LODs are one part of larger optimization work. They work best alongside:

  • sensible texture use
  • material control
  • good scene layout
  • culling where appropriate
  • mesh reduction or decimation in Blender or Maya

That means LODs are useful, but they are not the only performance tool.

Helpful follow-up pages

Final Advice

LODs are about spending detail where the player will notice it. The closest version gets the best quality. Farther versions get cheaper representations.

If you use LODs intentionally, especially on repeated or detailed objects, they can help a scene stay visually solid while using resources more efficiently.

Help! I do not know what to optimize first.

Start with the visible costs: geometry, materials, texture sizes, lighting, mirrors, effects, and anything active near spawn or busy social areas.

Help! The scene only slows down in VRChat.

Test with realistic avatars, platform targets, and active world systems. Editor-only testing can miss the costs that appear in real multiplayer use.

Related Navigation

References