Insights / Gaussian Splatting
Gaussian Splatting vs Photogrammetry: What's the Difference?
Gaussian splatting renders photoreal scenes in real time from 20–50 images; photogrammetry builds editable meshes from 100–500+. Cost and uses compared.
By Canopy Creative, Immersive experience studio, Melbourne. Published 14 May 2025, updated 1 October 2026.
The short answer: Gaussian splatting represents a scene as millions of soft 3D particles that render photorealistically in real time, while photogrammetry reconstructs a textured polygon mesh you can edit, 3D print or bring into CAD. Choose Gaussian splatting for VR, AR and web experiences where realism and speed matter; choose photogrammetry when you need precise, editable geometry.
Gaussian Splatting is quickly emerging as one of the most exciting breakthroughs in the world of 3D scanning and real-time graphics. For industries working in VR, AR, digital twins, and immersive storytelling, it promises faster workflows, more lifelike visuals, and real-time rendering performance unlike anything photogrammetry can achieve on its own.
But what exactly is Gaussian Splatting? How does it compare to traditional photogrammetry?
And when should you use one technique over the other?
In this comprehensive guide, we break down what Gaussian Splatting is, how it works, and how it stacks up against photogrammetry in terms of workflow, output quality, cost, and best-use scenarios.
What Is Gaussian Splatting?
Gaussian Splatting is a novel technique for rendering 3D scenes using millions of overlapping volumetric particles, or "Gaussians", each carrying color, transparency, position, and size data. These Gaussians replace the need for traditional polygon meshes by creating a soft, continuous representation of objects and environments directly from captured images.
This method evolved from research into Neural Radiance Fields (NeRFs) but improves drastically on rendering speed and visual quality, making it practical for real-time applications like VR, AR, and interactive 3D content.
Key Characteristics of Gaussian Splatting
- Uses point-based rendering, not mesh geometry.
- Delivers high-fidelity, cinematic-quality visuals.
- Can run in real-time on modern GPUs.
- Ideal for scanning environments, objects, and full scenes.
How Gaussian Splatting Works
- A camera (or set of cameras) captures a series of overlapping images.
- These images are aligned using camera tracking data or structure-from-motion techniques.
- An optimisation process places millions of 3D Gaussians throughout the scene and refines them until renders match the photos.
- Each Gaussian contains position, color, and light scattering properties.
- The scene is rendered in real time using a GPU-accelerated renderer.
The end result is a scene that appears volumetric and highly photoreal, without the need for polygon modeling or traditional texture mapping.
What Is Photogrammetry?
Photogrammetry is a proven method that reconstructs 3D geometry by analyzing multiple 2D photographs taken from different angles. It generates a dense point cloud, which is then converted into a 3D mesh with a photo texture overlaid.
Photogrammetry is widely used in architecture, archaeology, game development, and product visualization due to its ability to produce precise, editable 3D models.
Key Steps in Photogrammetry
- Capture hundreds of high-resolution images.
- Stitch and align images using photogrammetry software.
- Generate a dense point cloud.
- Convert point cloud into a mesh.
- Apply texture mapping for realism.
- Export to 3D file formats for further use in engines like Unity or Unreal.
Gaussian Splatting vs Photogrammetry Compared
While both techniques convert real-world imagery into 3D digital form, the approach, output format, and applications are fundamentally different.
Output: Gaussian Particles vs Polygon Mesh
| Feature | Gaussian Splatting | Photogrammetry |
|---|
| Format | Volumetric Gaussian particles | Polygon mesh with textures |
| Editability | Limited (non-mesh based) | Fully editable mesh |
| File Size | Smaller and GPU-optimized | Larger with detailed geometry |
| Output Type | Real-time renderer | Exportable 3D asset |
| Compatibility | Optimized for real-time rendering (VR, AR) | CAD, games, 3D printing, VFX |
Workflow and Performance Comparison
| Feature | Gaussian Splatting | Photogrammetry |
|---|
| Image Input Required | 20–50 images | 100–500+ images |
| Real-Time Ready | Yes | Not natively |
| Software | Custom or open-source AI tools | Agisoft, RealityCapture, Meshroom |
| Hardware Requirements | High-performance GPU | DSLR or mirrorless cameras, CPU/GPU |
Visual Quality Comparison
- Gaussian Splatting: delivers smooth lighting transitions, cinematic depth, and realistic parallax.
- Photogrammetry: provides surface precision and detail, useful for engineering and digital twins where exact geometry matters.
Advantages of Gaussian Splatting
- Real-Time Performance: Gaussian Splatting can be rendered in real time, making it perfect for VR, XR, and web-based interactive environments.
- Cinematic Visuals: Lighting, shadows, and materials appear soft and natural, similar to film-like rendering or volumetric photography.
- Minimal Input, Maximum Output: It requires far fewer images to build a complete 3D representation compared to traditional photogrammetry.
- No Mesh Cleanup: Because the output is point-based, there’s no need for UV unwrapping, decimation, or manual retopology.
- AI-Native and Scalable: Built for the AI generation of tools, it fits into automated pipelines and future neural graphics workflows.
Limitations of Gaussian Splatting
- Difficult to Edit: You cannot manipulate or animate individual components like you would in a mesh-based model.
- Incompatible With CAD or Print: Outputs are not suitable for 3D printing or engineering analysis.
- Experimental Ecosystem: Still a cutting-edge method with tools primarily in research or developer stages.
- Requires Technical Expertise: Most pipelines require Python, command-line tools, and ML model management.
Animation is starting to change: 4D Gaussian Splatting adds time to the technique so moving scenes can be captured and rendered in real time.
Gaussian Splatting vs Photogrammetry Cost
| Factor | Gaussian Splatting | Photogrammetry |
|---|
| Camera Equipment | Low (action cam or phone) | Medium to High (DSLR + lighting) |
| Software | Free (open-source) | $0–$1500 depending on license |
| Processing Time | Fast | Slow (image cleanup, mesh generation) |
| Editing Flexibility | Low | High |
| Real-Time Compatibility | Native | Not native (needs optimization) |
Best Use Cases for Gaussian Splatting
- Virtual Reality Scenes: Realistic, responsive worlds with low latency and lifelike depth.
- AR/VR Training Simulations: Lightweight assets for headset deployment with fast load times.
- Exhibitions and Museums: Real-time, cinematic captures of environments or installations.
- Digital Twins for Live Monitoring: Interactive spatial scans that run in real time for remote collaboration.
- Virtual Showrooms: Showcasing environments with rich visual depth and intuitive interaction.
When to Choose Photogrammetry Over Gaussian Splatting
- Precision modeling for CAD or 3D printing
- Objects with clear edges, labels, or fine structure
- Scenes needing export to traditional 3D engines
- Animation pipelines requiring character rigging
- Static architectural visualization or reverse engineering
How Canopy Creative Uses Gaussian Splatting
At Canopy Creative, we integrate Gaussian Splatting into our immersive design workflows for clients in education, entertainment, and commercial sectors. We use it to create ultra-realistic environments and interactive scenes that load instantly on VR headsets, WebXR browsers, or XR displays.
When projects require traditional deliverables, like printable 3D assets or animated characters, we blend in photogrammetry or LiDAR scanning.
Our hybrid approach ensures you're using the right tool for the right task, without compromising on performance or visual fidelity.
Gaussian Splatting Is the Future of Real-Time 3D Capture
As real-time 3D continues to evolve, Gaussian Splatting offers a glimpse into the future of how environments are digitized, rendered, and shared. While photogrammetry remains an essential tool for precision modeling, Gaussian Splatting delivers unmatched performance and visual realism for immersive experiences.
If you’re developing a VR training app, building a virtual showroom, or launching an interactive brand campaign, Gaussian Splatting is one of the most efficient, scalable, and visually impressive technologies available today.
Gaussian Splatting or Photogrammetry? Ask Our 3D Team
Tell us what you need to capture and where it will be shown. We'll recommend Gaussian splatting, photogrammetry or a hybrid. Talk to Our 3D Team
Gaussian Splatting vs Photogrammetry FAQs
Is Gaussian splatting better than photogrammetry?
It depends on the output you need. Gaussian splatting is better for photoreal scenes that must run in real time in VR, AR or a web browser. Photogrammetry is better when you need precise, editable mesh geometry for CAD, 3D printing, rigging or engineering.
How many photos do you need for Gaussian splatting?
Typically 20–50 overlapping images, compared with 100–500+ high-resolution images for photogrammetry.
Can you 3D print a Gaussian splat?
No. Gaussian splatting produces volumetric particles rather than a mesh, so the output isn't suitable for 3D printing or engineering analysis. Use photogrammetry or LiDAR scanning when you need a printable model.
What software is used for photogrammetry and Gaussian splatting?
Photogrammetry commonly uses Agisoft, RealityCapture or Meshroom. Gaussian splatting mostly uses custom or open-source tools, which often need Python and command-line experience.
Can Gaussian splatting and photogrammetry be used together?
Yes. Canopy Creative uses Gaussian splatting for real-time environments on headsets, WebXR browsers and XR displays, and blends in photogrammetry or LiDAR when a project also needs printable 3D assets or animated characters.