Reflection Environment

The Reflection Environment feature provides efficient glossy reflections in every area of the level. Many important materials like metals rely on having reflections in all directions, which the Reflection Environment provides. It is targeted toward consoles and mid spec PC, so it must run very fast. Support for reflections of dynamic objects or sharp reflections is supported but will require addational memory overhead.

Quick Reflection Environment Setup

To quickly get the Reflection Environment working in your project's levels, you will need to do the following:

1. Add a few lights to your level and build the lighting once as there must be some indirect diffuse lighting for the Reflection Environment to show up at all.

2. From the Place Actors panel in the Visual Effects tab, select and drag a Sphere Reflection Capture Actor into the level.

If you fail to see reflections in your level or your reflections are not as strong as you require you can try the following:

• Make sure that your Materials have a noticeable Specular and a low Roughness to better show reflections.
• Use the Reflection Override view mode to visualize what is being captured to gain a better idea of what values in your Materials need to be adjusted.

Setting Up a Level to use the Reflection Environment

The first step toward having good reflections is setting up diffuse lighting including indirect lighting through the use of lightmaps. The Lightmass page contains more info on how to accomplish this if you are unfamiliar with using it. Common errors preventing Lightmass indirect lighting from working after a build include but are not limited to the following:

• A shadow casting skybox.
• Lack of a LightmassImportanceVolume.
• Missing or incorrectly setup lightmap UVs
• Having Force No Precomputed Lighting set to True in the World Properties.

Since the level's diffuse color is what will be reflected through the Reflection Environment you will need to do the following for the best results.

• Ensure significant contrast between directly lit and shadowed areas.
• Remember that the bright diffuse lit areas are what will show up clearly in reflections.
• Darker shadowed areas are where the reflections will be most visible.
• Use the Lit viewmode together with the Specular show flag disabled to see the level as the reflection captures see it.

It is also extremely important to setup your level's Materials to work well with the Reflection Environment by keeping the following in mind.

• A flat, mirror surface will reveal the inaccuracies of combining cubemaps projected onto simple shapes.
• Curvy geometry or rough surfaces can both obscure these artifacts and provide convincing results.
• It is important to use detail Normal maps and specify some degree of roughness on Materials that will be used in flat areas as this will help them better show off reflections.

Place reflection captures in the areas that you want to have reflections. Try to place the sphere captures such that the part of the level you want to reflect is contained just inside their radius since the level will be reprojected onto that sphere shape. Try to avoid placing captures too close to any level geometry, as that nearby geometry will dominate and block important details behind it.

Glossy Indirect Specular

In technical terms, the Reflection Environment provides indirect specular. We get direct specular through analytical lights, but that only provides reflections in a few bright directions. We also get specular from the sky through a Sky Light, but that does not provide local reflections since the Sky Light cubemap is infinitely far away. Indirect specular allows all parts of the level to reflect on all other parts, which is especially important for Materials like metal which have no diffuse response.

Full lighting

The Reflection Environment works by capturing the static level at many points and reprojecting it onto simple shapes like spheres in reflections. Artists choose the capture points by placing ReflectionCapture Actors. Reflections update in real time during editing to aid in placement but are static at runtime. Projecting the captured level onto simple shapes gives approximate parallax for the reflection. Each pixel blends between multiple cubemaps to get the final result. Smaller ReflectionCapture Actors override larger ones, so you can refine reflection parallax accuracy in areas as needed. For example, you might place a capture at the center of a room and then refine the reflection by placing smaller captures at the corners of the room.

Materials with varying glossiness are supported by generating blurry mipmaps from the captured cubemaps.

However, just using the cubemap reflections on a very rough surface results in an overly bright reflection that has significant leaking due to lack of local occlusion. This is solved by reusing the lightmap data generated by Lightmass. The cubemap reflection is mixed together with the lightmap indirect specular based on how rough the material is. A very rough material (fully diffuse) will converge on the lightmap result. This mixing is essentially combining the high detail part of one set of lighting data (cubemaps) with the low-frequency part of another set of lighting data (lightmaps).

For this to work correctly, though, only indirect lighting can be in the lightmap. This means that only the indirect lighting from Stationary lights can improve the quality of reflections on rough surfaces. Static light types should not be used together with the Reflection Environment as they will put direct lighting in the lightmap. Note that this mixing with the lightmap also means that the map must have meaningful indirect diffuse lighting and that lighting must already be built to see results.

Reflection Capture Lightmap Mixing

When you use Reflection Capture Actors, UE4 mixes the indirect Specular from the Reflection Capture with the indirect Diffuse lighting from lightmaps. This helps to reduce leaking since the reflection cubemap was only captured at one point in space, but the lightmaps were computed on all the receiver surfaces and contain local shadowing information.

While lightmap mixing works well for rough surfaces, this method breaks down on smooth surfaces as reflections from Reflection Capture Actors no longer match reflections from other methods like Screen Space Reflections or Planar Reflections. Because of this, lightmap mixing is no longer applied to very smooth surfaces. A surface with Roughness value of 0.3 will get full lightmap mixing, fading out to no lightmap mixing by Roughness 0.1 and below. This also allows Reflection Captures and Screen Space Reflections to match better and make it harder to spot transitions between the two.

Lightmap Mixing and Existing Content

By default, lightmap mixing will be enabled which means it will affect existing content. In cases where you had reflections leaking on smooth surfaces, that leaking will be more apparent. To solve this, you can either place additional Reflection Capture Actors around the level to help reduce the leaking. Or you can revert to the old lightmap mixing behavior by going to Edit > Project Settings > Rendering > Reflections and then un-check Reduce lightmap mixing on smooth surfaces.

You can fine tune the amount of lightmap mixing that will take place by adjusting the following commands via the UE4 console.

• r.ReflectionEnvironmentBeginMixingRoughness (default = 0.1)
• r.ReflectionEnvironmentEndMixingRoughness (default = 0.3)
• r.ReflectionEnvironmentLightmapMixBasedOnRoughness (default = 1)
• r.ReflectionEnvironmentLightmapMixLargestWeight (default = 1000)

High Quality Reflections

While the default reflection quality settings strike a good balance between performance and visual quality, there could be instances where you want to achieve even higher quality reflections. The following sections describe the available methods for achieving high quality reflections.

High Precision Static Mesh Vertex Normal and Tangent Encoding

An important factor in achieving high quality reflections is how accurately the vertex normal and tangent can be represented. Very high density meshes may lead to adjacent vertices quantizing to the same vertex normal and tangent values. This can lead to blocky jumps in normal orientation. We added the option to encode normals and tangents as 16 bits per channel vectors which enables developers to make the trade off between higher quality and how much additional memory is used encoding vertex buffers.

To enable High Precision Static Mesh Vertex Normal and Tangent Encoding:

1. In the Content Browser, double-click on a Static Mesh to open it up in the Static Mesh Editor.

2. In the Static Mesh Editor, go the Details panel and expand the LOD0 option.

3. At the bottom of the LOD0, there is a section called Build Settings. Click on the small triangle next to Build Settings to expand the Build Settings options.

4. Enable the Use High Precision Tangent Basis option by clicking on the check box next to its name and then press the Apply Changes button to apply the new settings.

The viewport will automatically update to reflect the changes.

The quality of the reflection that is viewed is directly related to how densely tessellated the Static Mesh is. Static Meshes that have less tessellation will have more stretching artifacts in the reflection than Static Meshes that have more tessellation.

High Precision GBuffer Normal Encoding

Enabling the High Precision GBuffer Normal Encoding option will allow the GBuffer to use a higher precision Normal encoding. This higher precision GBuffer Normal encoding encodes the Normal vector into three channels with each channel having 16 bits per. Using this higher precision encoding allows techniques like Screen Space Reflections (SSR) to rely on high precision normals.

To enable High Precision GBuffer Normal Encoding:

1. In the Main Toolbar, select Edit > Project Settings to open the Project Settings.

2. In the Project Settings under the Engine section, click on the Rendering option and under the Optimizations section change the Gbuffer Format from Default to High Precision Normals.

   Keep in mind that this encoding requires increased GPU memory and enabling this will have a direct impact on your project's performance.

Since changing the GBuffer format does not require you to restart the Editor, you can quickly change between the different GBuffer formats to see the impact they will have on the reflection visuals. In the image below we can see how changing the GBuffer format from the Default to High Precision Normals changes the look and quality of the reflection.

Reflection capture shapes

There are currently two reflection capture shapes: sphere and box. The shape is very important because it controls what part of the level is captured into the cubemap, what shape the level is reprojected onto in reflections, and which part of the level can receive reflections from that cubemap (area of influence).

Sphere shape

The sphere shape is currently the most useful. It never matches the shape of the geometry being reflected but has no discontinuities or corners, therefore, the error is uniform.

The sphere shape has an orange influence radius that controls which pixels can be affected by that cubemap, and the sphere that the level will be reprojected onto.

Smaller captures will override larger ones, so you can provide refinement by placing smaller captures around the level.

Box shape

The box shape is very limited in usefulness and generally only works for hallways or rectangular rooms. The reason is that only pixels inside the box can see the reflection, and at the same time all geometry inside the box is projected onto the box shape, creating significant artifacts in many cases.

The box has an orange preview for the projection shape when selected. It only captures the level within Box Transition Distance outside this box. The influence of this capture fades in over the transition distance as well, inside the box.

Editing Reflection Probes

When making edits to Reflection Probes there are a number of different things that you must remember to do to ensure that you get the results you are after. In the following section we will cover what these things are and how you can make sure you are getting the best quality reflections in your projects.

Updating Reflection Probes

It is important to note that Reflection Probes are not automatically kept up to date. Only the following actions will automatically update the Reflection Probes place in a level.

• Loading a map.
• Directly editing a Reflection Capture Actor properties.
• Building the levels lighting.

If you make any other kind of edit to the level like change a light's brightness or move around level geometry, you will need to select a Reflection Capture Actor and click the Update Captures button to propagate the changes.

Using a Custom HDRI Cubemap in a Reflection Probe

Reflection Probes have the ability to not only specify which cubemap they should be using for reflection data but also what size that cubemap should be. Previously UE4 hard-coded the resolution of the cooked cubemaps reflection probes would use. Now developers can choose the resolution that best suits their needs based on performance, memory, and quality tradeoffs. Below is a comparison image that shows the difference between using the Captured Scene option versus the Specified Cubemap option.

To specify a custom HDRI Cubemap for your project's Reflection Probes to use you will need to do the following:

1. First, make sure that you have an HDRI Cubemap Texture available for use. If you do not have an HDRI Cubemap Texture in your project, one comes bundled with the Starter Content called HDRI_Epic_Courtyard_Daylight.

   You can move assets from one project to another one using the Migrate functionality.

2. Select a Reflection Probe Actor that has been placed in the level and in the Details panel under the Reflection Capture section change the Reflection Source Type from Captured Scene to Specified Cubemap

3. With the Reflection Probe still selected in the level, go to the Content Browser and select the HDRI Texture you want to use. Then in the Reflection Capture Actor, under the Reflection Capture drag the HDRI Texture from the Content Browser to the Cubemap input.

4. Press the Update Capture button to refresh the Reflection Capture Actor to use the new HDRI Cubemap Texture that was just specified.

Adjusting Reflection Probe Resolution

You can globally adjust the resolution of the HDRI Cubemaps that are used for the Reflection Capture Actors by doing the following:

1. Open up your Project settings by going to the Main Toolbar and then selecting Edit > Project Settings.

2. From the Project Settings menu go to the Engine > Rendering section and then look for the Textures option.

3. By adjusting the Reflection Capture Resolution option you can increase or decrease the size of the HDRI Cubemap Texture that was specified.

   

   Please note that the Cubemap Resolution will only work with numbers that are powers of 2 like 16, 64, 128, 256, 512, and 1024. When using values other than a power of 2, it will be rounded to the nearest accepted resolution value. Also, be extremely careful when using high resolution values as it will dramatically affect performance due to the GPU memory requirements.

The following images show what how the reflections will look when the Reflection Capture Resolution is set to 1, 4, 8, 16, 32, 64, 128, 256, 512 and 1024.

Adjusting Skylight Reflection Resolution

Like with the Reflection Probes, Skylights also have the ability to define and adjust the resolution of the HDRI cubemap that they use for reflections. To utlize this functionality in your UE4 project you will need to do the following:

1. From the Place Actors panel in the Lights tab, select and then drag a Skylight into your level.

2. Select the Skylight and in the Details panel under the Light section, change the Source Type from SLS Captured Scene to SLS Specified Cubemap.

3. Click on the drop down box in the Cubemap section and select an HDRI cube map from the list.

4. After the cubemap has been selected you can adjust its resolution by changing the value in the Cubemap Resolution input.

   Please note that the Cubemap Resolution will only work with numbers that are powers of 2 like 16, 64, 128, 256, 512, and 1024. When using values other than a power of 2, it will be rounded to the nearest accepted resolution value. Also, be extremely careful when using high resolution values as it will dramatically affect performance due to the GPU memory requirements.

Blending Multiple Reflection Probe Data

You can blend between multiple different cubemap reflections by providing the Reflection Capture Actors with different HDRI cubemaps. To accomplish this in your UE4 project all you need to is the following:

1. First, make sure that you have at least one Reflection Probe added to your level and that you have changed the Reflection Source Type to Specified Cubemap and input an HDRI Texture into the Cubemap input.

2. Duplicate or add a new Reflection Probe to the level and position / adjust its Influence Radius it so that part of it's yellow influence radius is intersecting with the first Reflection Probe.

3. Select the newly duplicated / created Reflection Probe Actor and in the Details panel under the Cubemap section change the HDRI cubemap to a different one.

4. With the Reflection Probe that was added / duplicated still selected, go to the Details panel in the Reflection Capture section and press the Update Captures button to update the reflection to use what was input in the Cubemap input.

5. If you select and move the Reflection Probe around the level you can get a better idea for how the two HDRI cubemaps will blend together.

Visualizing

The Reflection Override viewmode has been added to make it easier to see how well the reflections are set up. This viewmode overrides all normals to be the smooth vertex normal, and makes all surfaces fully specular and completely smooth (mirror like). Limitations and artifacts of the Reflection Environment are also clearly visible in this mode so it is important to switch to Lit periodically to see what the reflections look like in normal conditions (bumpy normals, varying glossiness, dim specular).

Some new show flags have been added which are useful for isolating down the components of the lighting:

Performance Considerations

The Reflection Environment cost is only dependent on how many captures influence the pixels on the screen. It is very similar to deferred lighting in this sense. Reflection captures are bounded by their influence radius and therefore they are culled very efficiently. Glossiness is implemented through the cubemap mipmaps so there is little performance difference between sharp or rough reflections.

Limitations

• Reflections through this method are approximate. Specifically, the reflection of an object will rarely match up with the actual object in the level due to projection onto simple shapes. This tends to create multiple versions of that object in reflections as many cubemaps are being blended together. Flat, smooth surfaces that cause mirror reflections will show the error very noticeably. Use detail normal maps and roughness to help break up the reflection and these artifacts.
• Capturing the level into cubemaps is a slow process which must be done outside of the game session. This means dynamic objects cannot be reflected, although they can receive reflections from the static level.
• Only the level's diffuse is captured to reduce error. Purely specular surfaces (metals) will have their specular applied as if it were diffuse during the capture.
• There can be significant leaking when there are different lighting conditions on both sides of a wall. One side can be setup to have correct reflections, but it will always leak into the other side.
• Due to DX11 hardware limitations, the cubemaps used to capture the level are all 128 on each side, and the world can have at most 341 reflection captures enabled at one time.