PolySpatial Render Texture Support

PolySpatial transfers RenderTextures to host platforms using an optimized path. On visionOS, this is performed by using a GPU blit to copy the contents of the RenderTexture to a texture provided by RealityKit. Transferring large numbers of RenderTextures at once and/or RenderTextures with large dimensions may still incur a performance penalty, however.

Formats

Formats supported for RenderTextures on visionOS are limited by the underlying DrawableQueue API. The only formats currently tested are R8G8B8A8_UNORM (the default RenderTexture format) and R16G16B16A16_SFLOAT. Other formats may or may not work; we recommend experimenting to find out.

Batch mode rendering

RealityKit apps for visionOS run in batch mode, which means that cameras do not render automatically. If you use a RenderTexture as the Output Texture of a camera, you may use the following script to ensure that the camera renders once per update.

using UnityEngine;
using Unity.PolySpatial;

public class BatchModeUpdateRenderer : MonoBehaviour
{
    Camera m_Camera;

    void Start()
    {
        m_Camera = GetComponent<Camera>();
    }

    void Update()
    {
        if (Application.isBatchMode && m_Camera)
            m_Camera.Render();
    }
}

Manual dirtying

If you update a RenderTexture outside of a call to Camera.Render (for example, by setting RenderTexture.active and invoking immediate mode rendering commands), PolySpatial will not automatically transfer the RenderTexture's new contents. Instead, you must manually dirty the RenderTexture using Unity.PolySpatial.PolySpatialObjectUtils.MarkDirty(renderTexture) each frame that it is updated. This will cause PolySpatial to send the new contents of the RenderTexture to the host.

using UnityEngine;
using Unity.PolySpatial;

public class SetRenderTextureDirty : MonoBehaviour
{
    public RenderTexture texture;

    void Update()
    {
        Unity.PolySpatial.PolySpatialObjectUtils.MarkDirty(texture);
    }
}