Skip to main content
Complete Ecosystem Guide — 2026

Gaussian Splatting : Real-Time 3D Capture

Gaussian Splatting (3DGS) is the fastest-growing 3D reconstruction method in 2026 — from smartphone apps to building-scale drone scanning. This guide covers every tool, workflow, and use case in the GS ecosystem: DJI Terra, Xgrids LCC, Polycam, Luma AI, Nerfstudio, and more.

The 3D scanning market is projected to reach $19-22 billion by 2033-2035 (13-16% CAGR from $5-6.7B in 2025), and Gaussian Splatting is driving a fundamental shift in how buildings, environments, and locations are captured and visualized.

12,100+
Monthly searches
60+ FPS
Real-time rendering
$2,250
Minimum project
Zero
US competitors

What Is Gaussian Splatting and How Does It Work?

Gaussian Splatting (3DGS) is a 3D reconstruction technique introduced by Kerbl et al. at SIGGRAPH 2023 that represents scenes as millions of overlapping 3D Gaussian ellipsoids. Unlike mesh-based photogrammetry (which creates triangulated surfaces) or Neural Radiance Fields (NeRF, which encodes scenes in neural network weights), GS uses explicit geometric primitives that can be rendered, edited, and streamed in real time.

Each Gaussian "splat" stores a position, 3D covariance matrix (size and orientation), color (via spherical harmonics), and opacity. A differentiable rasterizer projects these Gaussians onto screen space at 60+ frames per second — orders of magnitude faster than NeRF's per-pixel ray marching. The result is photorealistic quality at real-time speeds, with particularly strong performance on challenging materials: vegetation, glass, reflective surfaces, fine architectural details, and water features that traditional mesh reconstruction struggles to reproduce.

GS Accuracy in Context

Independent studies (plainconcepts.com, 2024) comparing Gaussian Splatting reconstructions to ground-truth LiDAR data found a mean geometric error of 7.82cm (standard deviation 11.49cm). By comparison, traditional photogrammetry with ground control points achieves 1-3cm accuracy, and professional LiDAR scanners like the Trimble X12 achieve ±2mm at 20m.

This makes Gaussian Splatting visualization-grade, not survey-grade. GS excels at creating photorealistic visual experiences — virtual tours, design visualization, virtual production environments, and public-facing presentations. For dimensional measurement, engineering compliance, and construction coordination, LiDAR and photogrammetry remain essential. The most powerful workflows combine both: LiDAR for the measurements, GS for the visuals.

Accuracy Comparison

Gaussian Splatting (3DGS) ~7.82cm mean error
Photogrammetry (with GCPs/RTK) 1-3cm
Xgrids L2 Pro (handheld SLAM) ±1-2cm relative
3DMakerPro Eagle (consumer) 2cm at 10m
NavVis VLX 3 (mobile SLAM LiDAR) ±5mm
Trimble X12 (terrestrial LiDAR) ±2mm at 20m

The Complete Gaussian Splatting Tool Ecosystem (2026)

Every tool, platform, and standard in the GS landscape — from enterprise drone pipelines to free smartphone apps.

Professional & Enterprise Tools

DJI Terra v5.2 (Flagship License)

$2,800-$4,400

Integrated drone-to-GS pipeline for DJI enterprise drones. Ingests aerial imagery, runs Structure-from-Motion (SfM), then Gaussian Splatting reconstruction. Outputs 3DTiles (for Cesium web viewers with level-of-detail streaming), PLY (Gaussian splats), and GeoTIFF. Processes approximately 500 images per hour. Supports batch processing and cluster deployment for city-scale projects up to 30,000 images. Requires Flagship license — Standard license does not include GS. Real-time 60 FPS preview during processing. Minimum 8GB GPU VRAM, recommended 24GB (RTX 4090).

Strengths: Integrated drone pipeline, georeferenced output, batch/cluster processing. Limitations: DJI drones only (vendor lock-in), desktop-only, aerial capture geometry only.

Xgrids LCC + Lixel L2 Pro

~$15,000-$25,000

Handheld SLAM-based GS system from Hong Kong-based Xgrids. The L2 Pro scanner captures 640,000 points/sec with 120m range, ±1-2cm accuracy, dual 48MP cameras, and 1TB onboard SSD. Xgrids LCC (Lixel CyberColor) processes scans into proprietary GS format with SDKs for Unreal Engine, Unity, and WebGL. Processing time: approximately 1 hour per 3-minute scan. Unique capability: LCC for Revit is the only GS-to-BIM plugin on the market. Xgrids is a contributing member to the Khronos glTF KHR_gaussian_splatting candidate specification. PortalCam (their newest product) is described as "the first true spatial camera."

Strengths: Handheld mobility, BIM integration, standards involvement. Limitations: Proprietary format ecosystem, high cost. Competitor Visualskies (UK) uses L2 Pro + LCC for film/VFX work.

Consumer/prosumer spatial scanner with native GS output. 200,000 points/sec, 80-140m range, 2cm accuracy at 10m, 48MP camera, hot-swap battery. Processed through RayStudio (not JMStudio) which outputs standard PLY (GS) + PLY (point cloud) + OBJ (mesh) — no proprietary lock-in. The newer Raven model (150K pts/sec, 80-100m, 12MP) released March 2026. Approximately 20x less accurate than professional scanners like the Trimble X12, but significantly more affordable and portable.

Strengths: Open output formats, portable, affordable. Limitations: Consumer-grade accuracy (2cm), limited range, not suited for building-scale projects.

Consumer & Smartphone Tools

Polycam

iOS / Android / Web — Freemium

Most user-friendly GS entry point. Supports photogrammetry, iPhone LiDAR, and dedicated GS modes. Cross-platform with cloud processing. Lowest barrier to entry for GS experimentation.

Luma AI

iOS / Web — Free cloud processing

Widely regarded as producing the best consumer-grade GS quality. Free cloud-based processing. Social sharing focus with direct Luma gallery embedding. Cloud-dependent with limited export control.

Scaniverse (Niantic)

iOS — Free

Free iPhone app from Niantic (Pokémon GO creators). Combines iPhone LiDAR with Gaussian Splatting for room-scale captures. Completely free with no subscription required.

KIRI Engine

Cross-platform web — Freemium

Web-based GS processing platform. Works from any device — upload photos from phone or camera, process in the cloud. Cross-platform accessibility without app installation.

Note: All consumer/smartphone GS tools are limited to walking-distance, ground-level capture. They cannot capture buildings from the air or achieve the coverage needed for commercial site documentation.

Developer & Desktop Tools

Nerfstudio / gsplat

Open Source (NVIDIA-backed) — Free

The leading open-source GS training framework. CUDA GPU required. gsplat is the GS-specific rasterization kernel. Active research community. Best for developers and researchers who want full control over the GS pipeline.

PostShot

Desktop — Free (Indie tier) / Studio

Commercial desktop GS application. Best quality from smartphone captures. GUI-based workflow for non-developers. Subscription model ($199/yr) has caused some community pushback, with users considering professional services instead.

SuperSplat

Open Source (PlayCanvas) — Free

Web-based GS editor. Clean, crop, merge PLY splat files directly in the browser. Essential post-processing tool for any GS workflow. No installation required.

SplatForge

Blender Addon — $49 one-time

Blender addon for GS integration. Handles up to 16 million splats in real-time within Blender's viewport. Enables GS editing alongside traditional 3D modeling workflows.

Standards & Interoperability (2026)

OpenUSD — Gaussian Splatting Support Ratified (April 2026)

The Universal Scene Description format now officially supports Gaussian Splatting data. This enables GS interoperability with NVIDIA Omniverse, Pixar tools, Apple Vision Pro, and the broader USD ecosystem. This is the most significant standards development for GS enterprise adoption.

Khronos glTF — KHR_gaussian_splatting Extension

Candidate specification for Gaussian Splatting within the glTF format. Xgrids is a contributing member to this specification. Once ratified, this will enable GS content in any glTF-compatible viewer, including web browsers, AR applications, and game engines.

SOG Format

Streamed level-of-detail format for large-scale GS scenes (10 million+ splats). Enables progressive loading similar to how 3DTiles handles large point clouds in Cesium.

Ecosystem Comparison Table

Tool Category Input Output Accuracy Price
DJI Terra Flagship Professional DJI drone photos 3DTiles, PLY, GeoTIFF Sub-cm visual $2,800-$4,400
Xgrids LCC + L2 Pro Professional Handheld SLAM scan Proprietary + SDKs ±1-2cm relative ~$15,000-$25,000
3DMakerPro RayStudio Prosumer Eagle/Raven scanner PLY, OBJ 2cm at 10m $3,398-$4,000
Polycam Consumer Phone photos/LiDAR PLY, USDZ Variable Freemium
Luma AI Consumer Phone photos PLY, MP4 Variable Free cloud
Nerfstudio/gsplat Open Source Any photos/video PLY Variable Free (CUDA GPU)
PostShot Desktop Photos/video PLY, SPLAT Variable $199/yr
SuperSplat Editor PLY splat files PLY (cleaned) N/A (editor) Free (PlayCanvas)

THE FUTURE 3D: Professional Gaussian Splatting for Buildings & Environments

Within the Gaussian Splatting ecosystem, THE FUTURE 3D occupies a specific niche: building-scale aerial GS for commercial, construction, heritage, and film applications. While consumer tools like Polycam and Luma AI handle room-scale captures, and Xgrids L2 Pro covers handheld interior scanning, THE FUTURE 3D uses drone-based aerial photogrammetry processed through DJI Terra Flagship to create GS reconstructions of entire buildings, campuses, and sites.

What makes this approach unique in the US market is the hybrid LiDAR + GS workflow. Every GS project includes survey-grade dimensional data from the Trimble X12 (±2mm accuracy) or NavVis VLX 3 (±5mm for interiors). The client receives both: GS photorealism for visualization and stakeholder presentations, plus LiDAR accuracy for engineering, compliance, and construction coordination. No other US-based provider currently offers this combination.

As of 2026, the only comparable international provider is Visualskies (London/Berlin/NYC), which uses DJI M4E drones paired with Xgrids L2 Pro handheld scanners. Visualskies has 48 ranked keywords globally. THE FUTURE 3D is among the first US companies to offer professional Gaussian Splatting as a named commercial service.

Our GS Capture Workflow

  1. 1
    Flight Planning & Aerial Capture — DJI Matrice 4 Enterprise (M4E) with Zenmuse P1 (45MP full-frame). 80%+ overlap for GS-optimal coverage. RTK/PPK for georeferencing.
  2. 2
    Ground-Truth LiDAR — Trimble X12 (±2mm) for exterior control. NavVis VLX 3 (±5mm) for interior scanning. Sets the dimensional accuracy baseline.
  3. 3
    GS Processing — DJI Terra Flagship runs SfM → Gaussian Splatting reconstruction. ~500 images/hour. Real-time 60 FPS preview during processing.
  4. 4
    Delivery — 3DTiles (Cesium web viewer), PLY (Gaussian splats), GeoTIFF, plus registered point clouds in E57/RCP/LAS from LiDAR. Client owns all data permanently.

THE FUTURE 3D does not scan individual objects, products, or props. Our GS services focus exclusively on buildings, environments, locations, and sites. For object-scale scanning, see consumer tools like Polycam, Luma AI, or 3DMakerPro Eagle.

Gaussian Splatting Pricing

Gaussian Splatting is a premium add-on to standard drone photogrammetry — it requires additional processing time, specialized software (DJI Terra Flagship), and GPU compute resources. GS pricing is set at 1.5x standard photogrammetry rates, reflecting the additional value and cost of the technology. GS processing is never free or included in standard photogrammetry deliverables.

Service Tier Standard Photogrammetry With Gaussian Splatting (1.5x)
Small Sites (up to 10 acres) $1,500-$3,000 $2,250-$4,500
Large Area (10-50 acres) $3,000-$10,000+ $4,500-$15,000+
Corridor/Linear $3,000-$10,000+ $4,500-$15,000+
Film Location Scout (new service tier) $3,000-$10,000
Film Full Set/Environment (new service tier) $10,000-$50,000

Minimum Gaussian Splatting project: $2,250. All projects include survey-grade LiDAR alongside GS — both deliverables are provided.

Pricing shown reflects average US rates. Actual costs vary by location based on local market conditions, regulations, and project logistics — both within the US and internationally. Get a custom quote

Our Equipment for Gaussian Splatting Projects

THE FUTURE 3D deploys purpose-built equipment for each phase of the hybrid LiDAR + GS workflow. Other ecosystem tools (Xgrids L2 Pro, Polycam, Nerfstudio) serve different scale ranges and use cases.

DJI Matrice 4 Enterprise (M4E)

Aerial Capture Platform

Enterprise drone for GS image acquisition. Paired with Zenmuse P1 (45MP full-frame) for maximum detail and overlap. RTK/PPK positioning for georeferenced output. THE FUTURE 3D's primary aerial capture tool.

DJI Terra Flagship

GS Processing Software

GS processing engine. Flagship license required for Gaussian Splatting (Standard does not include it). Outputs 3DTiles, PLY, GeoTIFF. ~500 images/hour, 30K max, 60 FPS preview. v5.0 (July 2025) → v5.2 current with L3 LiDAR fusion support.

Trimble X12

Survey-Grade LiDAR (±2mm)

Terrestrial laser scanner providing the dimensional accuracy baseline for GS projects requiring measurement. GS is visualization-grade; the X12 adds measurement-grade data. Paired on projects requiring both visualization and dimensional precision — clients receive both datasets.

NavVis VLX 3

Mobile Interior Scanning (±5mm)

Wearable mobile mapping system for interior spaces. 2.56M pts/sec, panoramic imagery. Used alongside exterior drone GS when interior documentation is needed. See our Mobile SLAM Scanning service for details.

Gaussian Splatting Software Compatibility

GS support varies widely across platforms. Web viewers and game engines lead adoption, while CAD/BIM software lags behind — with one notable exception. OpenUSD (April 2026) and Khronos glTF are the emerging interoperability standards.

Platform GS PLY 3DTiles Notes
Cesium (Web) Primary web delivery. TF3D standard.
Unreal Engine Via plugins (Volinga, Xgrids LCC SDK)
Unity Via Xgrids LCC SDK
Blender SplatForge addon, up to 16M splats
SuperSplat (Web) Open-source editor (PlayCanvas)
AutoCAD No GS support. Mesh/point cloud only.
Revit Xgrids LCC only GS-to-BIM plugin
Apple Vision Pro Via OpenUSD (April 2026)

Ready for Professional Gaussian Splatting?

THE FUTURE 3D is a leading US provider of professional GS services for buildings and environments. Hybrid LiDAR + GS from $2,250.

Gaussian Splatting FAQ

What is Gaussian Splatting? +
Gaussian Splatting (3DGS) is a 3D reconstruction technique that represents scenes as millions of overlapping 3D Gaussian ellipsoids rather than triangulated meshes or neural volumetric fields. Introduced by Kerbl et al. at SIGGRAPH 2023, it enables photorealistic real-time rendering at 60+ FPS from ordinary photographs — without the slow training times of Neural Radiance Fields (NeRF) or the geometric artifacts of traditional photogrammetry on complex materials like vegetation, glass, and water.
How accurate is Gaussian Splatting? +
Gaussian Splatting achieves visualization-grade accuracy with a mean geometric error of approximately 7.82cm (standard deviation 11.49cm), based on independent studies comparing GS reconstructions to ground-truth LiDAR data. This makes GS excellent for visualization, virtual tours, and design review — but it is not survey-grade. For engineering measurements, THE FUTURE 3D pairs GS visualization with LiDAR data from the Trimble X12 (±2mm at 20m) to deliver both photorealistic visuals and measurement-grade accuracy in a single project.
How much does professional Gaussian Splatting cost? +
Professional Gaussian Splatting processing is priced at 1.5x standard photogrammetry rates, reflecting the additional compute time and specialized software required. Small site GS projects (up to 10 acres) start at $2,250-$4,500. Large area GS mapping (10-50 acres) ranges from $4,500-$15,000+. Film location scouting scans range from $3,000-$10,000, and full film set/environment scans range from $10,000-$50,000. The minimum GS project is $2,250. Contact THE FUTURE 3D for a precise quote based on your project scope.
What equipment does THE FUTURE 3D use for Gaussian Splatting? +
THE FUTURE 3D captures aerial data for Gaussian Splatting using the DJI Matrice 4 Enterprise (M4E) drone with the Zenmuse P1 full-frame 45MP camera. The imagery is processed through DJI Terra Flagship, which includes native GS reconstruction. For ground-truth accuracy, we pair every GS project with LiDAR scans from the Trimble X12 (±2mm) or NavVis VLX 3 (±5mm for interiors). This hybrid approach — GS for photorealistic visualization, LiDAR for dimensional accuracy — is unique in the US market.
What is the difference between Gaussian Splatting and photogrammetry? +
Traditional photogrammetry reconstructs 3D meshes (triangulated surfaces) from photographs, achieving 1-3cm accuracy with ground control points. Gaussian Splatting also works from photographs but represents the scene as millions of colored Gaussian ellipsoids instead of mesh triangles. GS produces more photorealistic results — especially on challenging materials like vegetation, glass, reflective surfaces, and fine details — and renders in real-time at 60+ FPS. However, GS has lower geometric accuracy (approximately 7.82cm mean error vs 1-3cm for photogrammetry). The two techniques are complementary: photogrammetry for measurement, GS for visualization.
What is the difference between Gaussian Splatting and NeRF? +
Neural Radiance Fields (NeRF) represent scenes as neural networks that map 3D coordinates to color and density. NeRF training takes hours and rendering takes seconds per frame. Gaussian Splatting represents scenes as explicit 3D Gaussian primitives — no neural network required. GS trains faster (minutes vs hours), renders in real-time (60+ FPS vs seconds per frame), and produces editable 3D data rather than opaque neural weights. NeRF pioneered the concept; GS made it production-ready.
What software creates Gaussian Splats? +
The GS ecosystem includes several categories of software. Professional/enterprise tools include DJI Terra Flagship ($2,800-$4,400, drone aerial GS), Xgrids LCC (handheld SLAM-based GS with the only GS-to-BIM Revit plugin), and 3DMakerPro RayStudio (consumer spatial GS). Consumer apps include Polycam (freemium, iOS/Android/Web), Luma AI (free cloud processing, best consumer quality), and Scaniverse (free, Niantic). Developer tools include Nerfstudio/gsplat (NVIDIA-backed open-source), PostShot (commercial desktop), SuperSplat (open-source web editor), and SplatForge (Blender addon supporting 16M splats).
Can you use drone footage for Gaussian Splatting? +
Yes. DJI Terra Flagship processes drone imagery from DJI enterprise drones (M4E, Matrice 350, Mavic 3 Enterprise) into Gaussian Splats. The workflow is: plan flight → capture overlapping photos (80%+ overlap recommended) → import to DJI Terra → run GS reconstruction → export as 3DTiles, PLY, or GeoTIFF. THE FUTURE 3D uses the M4E with Zenmuse P1 (45MP) for GS capture, processing approximately 500 images per hour through DJI Terra Flagship.
What file formats does Gaussian Splatting use? +
Gaussian Splatting uses PLY as the primary file format — specifically, PLY files containing Gaussian splat parameters (position, covariance, color, opacity) rather than traditional point cloud PLY. DJI Terra also exports 3DTiles for web-based Cesium viewers with level-of-detail streaming. Emerging standards include OpenUSD (official Gaussian Splatting support ratified April 2026) and Khronos glTF with the KHR_gaussian_splatting extension candidate. The SOG format handles streamed LOD for 10M+ splat scenes.
Can Gaussian Splatting replace LiDAR? +
No. Gaussian Splatting and LiDAR serve different purposes and are complementary, not competitive. GS produces photorealistic visual reconstructions with approximately 7.82cm mean geometric error — ideal for visualization, virtual tours, design review, and virtual production. LiDAR (such as the Trimble X12 at ±2mm) produces survey-grade dimensional data for engineering, construction, and regulatory compliance. THE FUTURE 3D combines both: LiDAR for the measurements, GS for the visuals. This hybrid workflow delivers what neither technology can achieve alone.
Do you scan individual objects with Gaussian Splatting? +
No. THE FUTURE 3D's Gaussian Splatting services focus exclusively on buildings, environments, locations, and sites. We do not scan individual objects, products, props, or vehicles. Our drone-based GS capture is designed for building-scale projects: commercial properties, construction sites, heritage buildings, film sets (the environment, not props), and urban areas. For object-scale scanning, consumer tools like Polycam, Luma AI, and 3DMakerPro are well-suited alternatives.
Is Gaussian Splatting ready for commercial use? +
Yes. Gaussian Splatting has moved beyond research into commercial production tools. DJI Terra v5.0+ (July 2025) added native GS processing for enterprise drones. Xgrids LCC provides commercial GS with BIM integration. OpenUSD ratified GS support in April 2026, enabling interoperability with NVIDIA Omniverse, Pixar tools, and Apple Vision Pro. Khronos glTF has a candidate GS extension. The technology is production-ready for visualization, virtual tours, virtual production, and design review workflows.
Who are THE FUTURE 3D's competitors in Gaussian Splatting? +
As of 2026, there are zero US-based companies offering professional Gaussian Splatting as a named commercial service. The only international competitor is Visualskies (London/Berlin/NYC), which uses DJI M4E drones combined with Xgrids L2 Pro handheld scanners and has 48 ranked keywords. THE FUTURE 3D is among the first US companies to offer building-scale professional GS services, combining DJI Terra Flagship aerial GS with Trimble X12 survey-grade LiDAR accuracy — a hybrid approach no other provider currently offers.
How long does Gaussian Splatting processing take? +
DJI Terra Flagship processes approximately 500 images per hour for Gaussian Splatting reconstruction. A typical small building (200-500 images) completes in under 2 hours. A large commercial site (2,000-5,000 images) takes 4-10 hours. City-scale projects (up to 30,000 images) may take 1-3 days with cluster processing. Xgrids LCC processes approximately 1 hour per 3-minute handheld scan. Consumer apps like Polycam and Luma AI process in minutes to hours depending on cloud queue and scene complexity.
What industries benefit from Gaussian Splatting? +
Architecture (photorealistic design visualization and client presentations), real estate (immersive property marketing and virtual tours), construction (progress documentation with visual context), heritage preservation (cultural site documentation for museums and research), film and virtual production (scan-to-LED-wall environments, digital backlots), gaming (photorealistic environment capture for level design), and urban planning (city-scale visualization for public engagement and development review).
Does Gaussian Splatting work with CAD and BIM software? +
Direct GS support in CAD/BIM software (AutoCAD, Revit, SolidWorks) is currently limited. The one exception is Xgrids LCC, which offers the only GS-to-BIM plugin for Autodesk Revit. For other CAD/BIM workflows, GS files must be converted to point clouds or meshes first. GS has strong support in web viewers (Cesium via 3DTiles), game engines (Unreal Engine and Unity via plugins and Xgrids LCC SDKs), 3D editing tools (Blender via SplatForge addon, SuperSplat web editor), and emerging standards (OpenUSD, glTF).
1-Hour Response Time

Ready to Capture Your Space in Stunning 3D?

Join 300+ satisfied clients nationwide and internationally who trust THE FUTURE 3D for professional 3D scanning, Virtual Tours, and Digital Twin solutions.

No Hidden Fees
100% Satisfaction Guarantee
Licensed & Insured
Nationwide + International
Licensed & Insured
1hr Response