No capture hardware required
Skip multi-view photo walks and LiDAR sessions. Start from text, one reference image, or one short video clip.
Gaussian Splat · Generate, don't capture
Gaussian Splat Generator
Create explorable 3D Gaussian Splat scenes from a prompt, one image, or a short video—no LiDAR scan or multi-view capture required. Walk the result in Topview Canvas and send screenshots back to your image and video pipeline.
01 Before
Image → Gaussian Splat
02 After
Marble 1.1 Plus
Optional: add one image or one video (up to 30s). Image and video cannot be used together.
What is Gaussian Splatting?
3D Gaussian Splatting represents a scene as many oriented ellipsoids instead of a traditional polygon mesh. That makes free camera movement through the space feel natural. On Topview, splats are generated from AI inputs rather than from scanning a real location, so you can concept environments without a capture-first workflow.
Why generate splats instead of capturing them
Built for creators who need navigable scenes inside a Canvas workflow, not a photogrammetry export pipeline.
Explorable, not a flat wrap
Move through the splat in three dimensions with keyboard and mouse, then frame the shot you want.
Screenshot back to Canvas
Capture viewpoints in common aspect ratios and drop new image nodes beside your 3D World card for downstream creation.
Same Marble-powered pipeline
Gaussian Splat worlds on Topview use Marble models inside Canvas—one workflow for concepting and capture.
How to generate a Gaussian Splat scene
From prompt to explorable splat in Canvas.
01
Open 3D World in Canvas
From the Creation sidebar, open the 3D World dock on your Infinite Canvas.
02
Describe or reference the scene
Write a prompt and optionally add one image or one video. Choose Marble 1.1 or Marble 1.1 Plus, review credits, then generate.
03
Enter the splat and capture
Open full-screen exploration, move through the Gaussian Splat world, and screenshot the framing you need back onto the canvas.
Capture-first splatting vs AI generation
Both produce splats—Topview focuses on generation and Canvas integration.
Capture-first splatting workflows
- Typically start by photographing or scanning a real location from many angles.
- Strong when fidelity to an existing place is the goal.
- Often focus on reconstructing an existing place for viewing or downstream delivery.
Topview Gaussian Splat generation
- Starts from text, one image, or one short video—no on-site capture session.
- Best for concept environments, branded spaces, and shot planning.
- Explore inside Canvas and capture frames into your image and video pipeline.
Related 3D tools on Topview
Frequently asked questions
Answers about Gaussian Splat generation, capture vs AI workflows, and Topview Canvas.
No. Topview generates Gaussian Splat scenes from text, an image, or a short video. That is different from capture-first apps that require many photos or LiDAR scans of a real place.
Capture-first workflows usually begin with a real-world scan or many photos. Topview generates splats with AI inside Canvas so you can concept spaces without an on-site capture session, then explore and screenshot results.
No. Topview outputs explorable Gaussian Splat worlds for navigation and screenshot capture inside Canvas—not downloadable mesh files for printing or game engines.
A text prompt, one image (PNG / JPG / WebP), or one video (MP4 / MOV / WebM up to 30 seconds). Text can combine with either media type, but image and video cannot be used together.
It looks like a soft, detailed 3D scene you can move through freely. It is not a single 360° panorama wrapped around you—it has depth you can navigate.
Topview offers Marble 1.1 for standard worlds and Marble 1.1 Plus for Larger World generation. Credits are shown in Canvas before you generate.
In the full-screen viewer, set field of view and aspect ratio, then capture. PNG image nodes appear next to your 3D World card for further editing or generation.
Yes. Panoramas with at least a 2:1 aspect ratio can serve as the sole media input without additional text, depending on your scene needs.
No. Each generation accepts text, one image, or one short video. You can combine text with one media input, but image and video cannot be used together.
Ready to generate your first splat scene?
Open Topview Canvas, describe a space or add one reference, and explore the Gaussian Splat result in full screen.
Text, image, or short video · Explore and capture in Canvas