Probe lighting

This tutorial adds a tested lighting field to a room, then refines a smaller region where the lighting changes quickly. You will use a movable object to check the result rather than judging only the probe markers.

Complete First bake first. Keep the room geometry and lights unchanged while comparing coverage settings.

1 Choose a test receiver and route

Add a small sphere or cube with Mobility: Movable and a simple Default Lit material. Plan a short route through the room: a bright area, a shaded corner, a doorway, and the edge of the intended coverage.

Use ordinary lit materials for this first test. You do not need to add a Spectral Sampler to every material.

For mobile, confirm Support GPUScene on Mobile, restart after renderer changes, and follow the platform requirements. A desktop preview alone does not verify the mobile material path.

2 Fit the main field

1. Add Spectral Probe Volume from Place Actors > Spectral.
2. Position and resize its box to cover the complete test route, including the height occupied by the receiver.
3. Leave Enabled on and Movable off for a fixed room.
4. Start with Voxel Size: 100 cm, Priority: 0, and Fade Distance: 50 cm, then judge them against the scale of your scene.
5. Use Visualize Probes to inspect the planned grid.

You can use Scale To World as an initial fit, but review the result. It fits eligible scene geometry, not the path of your character, and can include more space than you need.

Checkpoint: markers cover the intended receiver positions. A visible box or marker grid does not mean lighting has been baked yet.

3 Bake and establish a baseline

Run Baker > Bake Lighting and let it finish. Save the level and lighting data. Check that the volume reports Baked, then inspect it with Visualize Global Probes.

In Visualize > Volume Visualizations, use Global Visualization Scale to adjust marker size without changing probe density. Keep Hide Intersecting Probes off while inspecting red intersection markers, then enable it if they obstruct your view of the rest of the field. Hiding them does not fix the intersections or change the bake.

Turn visualization off and move the receiver along the route. Use a simple movement test in Play mode if you already have one, or reposition the Movable test actor in the editor. Keep the lights fixed; the purpose is to observe the existing field.

Look for lighting changes that correspond to the room, rather than sudden jumps at arbitrary positions. Record any problem location before changing settings.

4 Add detail only where it helps

If the transition near a doorway or small bright opening is too broad, first check whether the grid is too sparse there.

1. Add a second, smaller Spectral Probe Volume around the problem region.
2. Use a smaller Voxel Size, such as 50 cm for a comparison against the 100 cm main field.
3. Keep the dense box focused on the doorway and nearby receiver positions.
4. Bake again, save, and repeat the same movement route.

Halving spacing in three dimensions can create roughly eight times as many samples for the same bounds. Do not apply the dense setting across the entire level just to fix one doorway.

Density and sampling are differentNote

Smaller Voxel Size adds positions where lighting is recorded. More bake samples improve the estimate at those positions. If the field is spatially too coarse, increasing GI Samples alone does not add the missing positions.

5 Control overlaps and boundary fades

At equal Priority, Spectral favors the denser field. Leave both fields at 0 if that matches your intention. If a particular volume must win regardless of density, give it a higher Priority and use Show Priority Next To All Volumes to inspect the arrangement.

Move the receiver across both the overlap and the outer edge. Fade Distance extends the field's transition beyond the box toward fallback lighting. Increase it modestly if the boundary is abrupt, but do not use a very broad fade to disguise missing coverage.

Priority does not improve a bad bake, and fade does not add probes outside the captured grid. If the fade reaches through an unrelated room, adjust the bounds and layout.

6 Diagnose a mismatch

What you see	Next check
No useful change across the room	Confirm valid baked data, active coverage, and that the chosen material uses the supported probe path.
A sudden change at a doorway	Inspect overlapping priorities, bounds, and spacing on both sides.
Bright light leaking across a wall	Check geometry and grid placement; compare a focused denser field before increasing brightness.
A smooth but overly broad transition	Compare smaller spacing. Interpolation cannot restore unrecorded spatial detail.
Markers hide everything	Reduce Visualization Scale. It does not change the bake.

Under Features > Probes, compare Bilinear and Bicubic interpolation only after the field layout is sound. This is a live sampling choice, not a new capture.

7 Optional: explicit material sampling

For an Unlit material that needs custom lighting, use Spectral Sampler. Multiply the intended surface color by its Normal output and connect the result to Emissive Color.

Test this separately from the Default Lit receiver. The sampler returns white outside coverage, so a fully colored Unlit object is not proof that it found valid baked probes.

8 Save and check the target

Refresh Memory > Current Level and compare probe count and payload before and after adding the dense region. Save the scene and test the same route in the target build.

Changing bounds, Voxel Size, or scene lighting requires rebaking. Interpolation, priority, visualization, and fade tuning do not capture new illumination.

Moving a field is not recapturing itWarning

Enable a volume's Movable option only when its saved lighting should travel with it during play. It will not learn the lighting of its new surroundings. For reusable Blueprint lighting, follow Lightmapped prefabs. Duplicating an ordinary probe actor discards its bake and requires a new one.