Skip to main content

Rendering Images

SplatTransform can render a splat scene to a lossless WebP image from a camera view you choose. Give the output a .webp extension and the scene is rasterized on the GPU instead of being written as splat data. Use it for thumbnails, previews, panoramas and camera-animation frame sequences.

Rendering needs a GPU: it does not run with -g cpu. See Which Features Need a GPU and, for servers, the Docker Backend guide.

Setting Up the Camera​

By default the camera sits at 2,1,-2, looks at the origin with a 60° vertical field of view, and renders 1280x720 over an opaque black background:

# Default 1280x720 render
splat-transform input.ply view.webp

# Custom camera and resolution
splat-transform input.ply view.webp \
    --camera-pos 2,1,-2 --camera-target 0,0,0 \
    --camera-fov 50 --resolution 1920x1080

# Transparent background
splat-transform input.ply view.webp --background 0,0,0,0

--camera-up sets the world up vector (default 0,1,0). --camera-near sets the near clip distance (default 0.2, matching the reference 3DGS rasterizer); splats closer to the camera than this are culled.

360° Panoramas​

--projection equirect renders a full 360°×180° equirectangular panorama from --camera-pos, with --camera-target setting the view direction at the center of the image. The resolution must be 2:1 (default 2048x1024), and --camera-fov must be omitted:

# 360° equirectangular panorama from camera position
splat-transform input.ply pano.webp \
    --projection equirect --camera-pos 0,1,0 --camera-target 0,1,1

Depth of field is not available with this projection.

Depth of Field​

--f-stop enables defocus blur, treating the camera as a lens with that aperture: smaller numbers give stronger blur. The focus plane defaults to the distance to --camera-target; set --focus-distance to focus elsewhere.

# Defocus blur (focus on camera-target, f/2.8 aperture)
splat-transform input.ply view.webp --f-stop 2.8

# Smoother aperture sampling for strongly defocused edges
splat-transform input.ply view.webp --f-stop 1 --dof-samples 64

# Defocus with explicit focus distance and a smaller world scale
splat-transform input.ply view.webp \
    --f-stop 2.8 --focus-distance 3 --sensor-size 0.1

An f-stop only means something relative to a sensor size. --sensor-size is the vertical sensor height in world units, and its default of 0.024 is a 35mm full-frame sensor when one world unit is one meter. Scale it with your scene's units: 0.24 if a world unit is a decimeter, 24 if it is a millimeter.

Each image averages --dof-samples views across the aperture (default 32). More samples reduce sampling artifacts in strongly defocused areas, at the cost of more render passes.

Motion Blur​

Setting --camera-pos-end enables camera motion blur. The camera moves from its start pose (--camera-pos, --camera-target, --camera-up) at shutter open to its end pose at shutter close, and the image averages renders at --motion-samples instants across the shutter. --camera-target-end and --camera-up-end default to the start values.

# Camera motion blur (dolly from start to end pose over a 180° shutter, 16 instants averaged)
splat-transform input.ply view.webp \
    --camera-pos 2,1,-2 --camera-pos-end 3,1,-2 \
    --shutter 0.5 --motion-samples 16

--shutter is the fraction of the start→end movement that is averaged, centered on its midpoint: 1.0 covers the full movement and the default 0.5 is a 180° shutter. Rendering cost grows with --motion-samples; too few samples show as discrete copies wherever the motion between instants exceeds a couple of pixels.

Combining Effects​

Depth of field and motion blur resolve visibility separately for each sample, then average in linear light before encoding the final image. Reconstructed splat colors are treated as sRGB; transparent outputs use premultiplied linear color during averaging. When both effects are enabled, each shutter instant uses --dof-samples aperture views, so the cost multiplies.

Camera Animations​

--camera-track renders a camera animation as a numbered frame sequence instead of a single view. The track can come from a SuperSplat editor project, a viewer settings settings.json with animTracks, or a plain frame list. A saved .ssproj is a ZIP archive, so extract it first and pass the extracted directory or its document.json. A frame list looks like this:

{
    "frameRate": 30,
    "frames": [
        { "position": [2, 1, -2], "target": [0, 0, 0], "fov": 60 },
        { "position": [2.1, 1, -1.9], "target": [0, 0, 0], "fov": 60, "up": [0, 1, 0] }
    ]
}

Editor projects and viewer settings are evaluated the way the editor and viewer play them back. A frame list is interpolated linearly between entries; frameRate defaults to 30, frames without a fov use --camera-fov, and frames without an up use --camera-up. Each frame is written as <name>.NNNN.webp, numbered by its frame index in the track.

# Render every frame of an extracted SuperSplat project's camera animation
# (writes view.0000.webp, view.0001.webp, ...)
splat-transform scene.ply view.webp --camera-track scene-project/

# Render frames 0-47 at 1920x1080, motion-blurred over half of each frame interval
splat-transform scene.ply view.webp --camera-track track.json \
    --frames 0-47 --resolution 1920x1080 --shutter 0.5

The track replaces --camera-pos and --camera-target, and its target is also the focus point for depth of field. Motion blur for a track comes from --shutter alone: it is off by default, and setting it blurs each frame over that fraction of the frame interval. --camera-pos-end cannot be combined with a track.

See Also​