Render passes let you bypass the default render pipeline in Spark AR Studio to produce multiple rendered results in your AR effect. While they’re a more advanced feature to work with, they give greater flexibility in what you’re able to create.
A render pass is a computation that uses a combination of textures and numerical parameters to produce another texture. Connecting the result to another render pass patch adds further complexity.
Use them to render scene objects, textures and visual shaders. They’re particularly useful for applying different visual shader treatments to separate scene elements or to the final camera feed.
|Scene Render Pass|
Renders a scene object and all of its children. Create this patch using the menu in the Patch Editor.
|Shader Render Pass|
Renders one or more visual shaders or textures, creating complex textures formed of multiple images.
|Delay Frame Patch|
Use the Delay Frame patch to reuse a render pass from the previous frame in the current frame.
|Face Extraction Render Pass|
Use the Face Extraction Render Pass patch to add the face tracker texture to effects that use render pass patches.
Use the Face Distortion patch group to add face distortion to effects with render pass patches.
To set up a custom render pipeline with render passes you’ll first need a patch representing the Screen Output property of the Device scene object. You can either:
Adding the Screen Output patch individually
To create this patch:
A consumer patch labeled Screen Output will be added to the Patch Editor:
What’s rendering is now driven by patches. Because there are no patches connected to the Screen Output patch the Simulator video will be replaced by a blue screen:
This is patch graph that includes a Scene Render Pass patch, creating a complete render pipeline.
To create this graph:
A graph will be added to the Patch Editor, containing a:
The camera texture patch is connected to the Background input in the Scene Render Pass patch. This renders the camera texture across the background of the scene.
The patch representing the device is connected to the Scene Object input port, rendering all children of the Device in the Scene panel over the input in the Background port.
In the image below we’ve added the default render pipeline to a scene featuring 3D text. The 3D text renders over the camera texture:
Without the camera texture patch the background of the scene is black by default:
To edit the background color, adjust the Default Color value in the patch.
The output of the Scene Render Pass patch is an image. The render size of the image can be set using the Size values in the Scene Render Pass patch.
You can’t use the face tracker texture with render passes. You should use the Face Extraction Render Pass instead.
Scene Render Pass patches won’t render an alpha channel.
Using a custom render pipeline will impact how you arrange your scene hierarchy, so plan your effect and rendering strategy in advance.
As with any AR effect, test early and iterate, adding complexity as you go.
Because large render pipelines can impact performance, it’s important to test on a range of devices.
Building your graph
Before you connect any type of render pass patch to another, you can see a preview of how your effect will look after connecting the patches. The preview is hidden by default, open the preview by clicking on the arrow on the bottom right of the patch.
In the example below we’re changing the Multiplier value in the Gaussian Blur patch to adjust the output of the Shader Render Pass patch it’s connected to. This causes the preview image to become more or less blurry
Visually debug your graph by frequently moving the Screen Output patch, connecting it at different stages.
To improve efficiency, take a non-linear approach — rather than render something again, recycle it from an earlier stage of your visual shader graph.