Render Modes#
Camera sensors in ovrtx support three render modes, each offering a different trade-off between image quality and performance. The render mode is set per RenderProduct using the omni:rtx:rendermode attribute.
def Scope "Render" {
def RenderProduct "Camera" {
int2 resolution = (640, 480)
rel camera = </World/Camera>
rel orderedVars = [<LdrColor>]
# Select path-tracing render mode
token omni:rtx:rendermode = "PathTracing"
def RenderVar "LdrColor" {
string sourceName = "LdrColor"
}
}
}
Available Modes#
Mode |
|
Description |
|---|---|---|
Real-Time Path-Tracing |
|
Full path-traced rendering with real-time denoising. Produces high-quality images with accurate global illumination, reflections, and shadows. This is the default mode and the best choice for most applications requiring visual fidelity. |
Path Tracing |
|
Progressive path tracing that accumulates samples over multiple frames for ground-truth-quality rendering. Best suited for offline or reference-quality workflows where convergence time is acceptable. |
Minimal |
|
Lightweight rasterization-based rendering with minimal GPU cost. Use this when you need maximum throughput and do not require path-traced lighting – for example, segmentation masks, bounding-box visualization, or high-FPS reinforcement learning loops. |
Choosing a Mode#
Default to Real-Time Path-Tracing for sensor simulation, synthetic data generation, and any workflow that requires physically accurate lighting.
Use Path Tracing when you need converged, reference-quality images and can afford to accumulate samples across frames.
Use Minimal when rendering throughput matters more than visual quality – for example, thousands of environments stepping in parallel for RL training.
Different RenderProducts in the same scene can use different render modes. For example, one camera could use Real-Time Path-Tracing for RGB output while another uses Minimal for fast segmentation.
See Outputs for the available outputs per render mode.