Overview
In Q2 2026, Site Scan for ArcGIS is focused on one goal: helping teams move from capture to insight faster, especially when projects get very big or need extreme detail. These upcoming releases deliver two highly requested capabilities:
- Large Mission Processing: Confidently process today’s massive drone captures without breaking missions into smaller jobs. With increased cloud processing capacity and clearer guardrails, you can deliver large, high‑resolution datasets faster and more predictably, so you focus on using your data, not managing its limits.
A mesh of Downtown Zurich and surrounding areas processed in a single mission on quality Ultra. Credit: Wingtra
- Gaussian Splat Layer Creation: Create highly photorealistic 3D layers that preserve fine detail (including complex surfaces like vegetation and thin structures) to improve visualization and communication with stakeholders. Gaussian splat generation will be available later this quarter.
Gaussian splat layer processed in Site Scan. Credit: ShareUAV
Gaussian splats retain fine detail on vegetation and thin structures.
Large Mission Processing
As drone mapping programs scale up, image counts can quickly outgrow old “safe” guidance, leading to the need split up missions, failed jobs, delays, and uncertainty about what’s supported.
With large mission support, Site Scan increases the maximum mission size and adds clearer expectations and guardrails so you can plan captures, split work intelligently, and run processing with more predictability.
Custom subscription: up to 300 GP or 10,000 images per mission
Custom subscription organizations can upload, merge, and process missions up to 300 GP (gigapixels) or 10,000 images, whichever comes first.
Named-user: up to 150 GP or 5,000 images per mission
Named-user organizations can upload, merge, and process missions up to 150 GP (gigapixels) or 5,000 images (2× the prior 2,500-image supported limit). These limits are now more strictly enforced to protect processing reliability. If your typical capture regularly exceeds this size, talk with your account manager about options for higher mission limits.
Tips for processing large missions successfully
With larger datasets, failures can be more expensive: you might not discover an issue until a job has been running for hours (or days). A few up-front checks can materially improve success rates and help you avoid avoidable reprocessing. Use the tips below to increase the likelihood your mission completes successfully, especially when you’re near the supported size limits.
Check the input data quality
- Ensure overlap and sidelap follow recommended best practices for your survey area. Too little overlap can cause gaps in the outputs or failures; unreasonably excessive overlap and sidelap can increase the chances of failures due to out of memory or out of disk, even missions that are under the size limit.
Example of an overlap diagram from the processing report with good coverage. - Ensure the dataset consists of a cohesive block of images. Disparate or non-adjacent image blocks are not intended to be processed together in one mission and can lead to incomplete results or failures.
- Ensure images are sharp, have consistent lighting, and do not have artifacts such as motion blur.
Choose the right processing settings for your needs
- Quality Ultra can produce excellent results but may not be necessary or appropriate for all large datasets. Ultra significantly increases processing time, thus increasing the chance of timeouts or other failures if the mission is already near the supported limit. Quality High provides the best balance between resolution, reliability, and processing time.
- The reality engine can produce highly detailed 3D photorealistic models in the mesh (and soon, Gaussian splats), but enabling 3D outputs, like Ultra, significantly increases the processing time. If 3D outputs are not needed for your mission, disable them in the processing menu.
Prepare your mission before processing
- Use a mission area boundary and check the boundary extent before processing. This feature works best when the boundary is close to the images’ extent.
The default mission area can sometimes leave swaths of empty space without sufficient coverage. 
Modify the mission area to reduce these gaps and improve the results. - If your dataset contains water, use a water body boundary. You might need to process once first to see the relative location of the water on your True Ortho before tracing a boundary.
- If you are planning to use GCPs with your mission and your GCPs are standard checkerboards or crosses, upload your GCPs ahead of time and enable Automatic GCP Detection in the project settings.
Automatic GCP Detection setting is located in the Project Settings.
Check your project coordinate system
- Ensure your project’s horizontal and vertical coordinate systems are compatible with each other. This will help avoid failures caused by missing transformations. This is especially important if you are planning to process with GCPs, or if you plan to enable the mesh, as Site Scan meshes are transformed from your project coordinate system to WGS84 / EGM96.
If a failure occurs
- Avoid reprocessing repeatedly with the same settings.
- Check if a processing report was created by opening the Export menu.
- If you don’t see one, the failure occurred during the adjustment step. This means that changing settings such as Quality and 3D products (mesh, Gaussian splat) shouldn’t have an impact, as these settings apply only to the reconstruction step.
- If you do see one, look for signs of other issues such as low overlap or bad camera calibration.
- Connect with Esri Technical Support so that we can investigate the issue.
Gaussian splat layers
Expected to be available in May 2026, Site Scan will support the creation of Gaussian splats* (see important limitations below). Gaussian splats are a photorealistic 3D layer that can preserve fine detail and improve stakeholder communication, particularly in scenes that are traditionally challenging for meshes, such as vegetation, thin structures, and complex geometries.
Gaussian splat generation runs in Site Scan’s hosted cloud environment, so you get consistent processing without needing to size local workstation hardware for peak workloads.
If you’ve used the mesh workflow before, Gaussian splats will feel familiar. Here’s what to expect:
- In the processing menu, check the box Generate Gaussian splat layer.
- Once processed, view your Gaussian splat by opening the mission’s renamed 3D view (formerly Mesh view).
The redesigned 3D view includes both the mesh and Gaussian splat layers. The Layer Offset applies to both, and measurements are located in a new tab header. - If needed, apply an offset so the Gaussian splat aligns with the 3D terrain basemap. The offset is for visualization only; it does not impact the layer itself.
- Optionally, hide the 3D terrain basemap for a more focused viewing experience.
While the 3D terrain basemap can provide useful geographic context, sometimes it can be a distraction if you are only focused on the layer. - Share your Gaussian splat layer to ArcGIS Online by clicking Share to ArcGIS in the mission’s lower left toolbar. Sign into your ArcGIS Online account and under 3D Data Products enable Gaussian Splat Layer.
- Export your Gaussian splat layer by clicking the Export option in the mission’s lower left toolbar.
Gaussian splat limitations in Site Scan
- Gaussian splats are available to Custom Subscription organizations hosted in the U.S. Site Scan region (global default for most users: sitescan.arcgis.com) or India region (sitescan.arcgis.in). It is not available in the E.U. Site Scan region (sitescan-eu1.arcgis.com) currently due to cloud infrastructure limitations. If you’re unsure which region you’re in, or you need access, check with your account manager for eligibility and options.
- Publishing Gaussian splat layers to ArcGIS Enterprise is not supported.
- Creating Sharelinks to a 3D scene that includes a Gaussian splat layer is not yet supported.
Up next
We’re continuing to expand what the reality mapping capabilities of ArcGIS offer, without adding complexity for field teams. As you try large mission and Gaussian splat processing, we’d love feedback on what’s working and where you hit friction (processing time, output quality, publishing workflows). Share your experience on Esri Community, or route details through your account team or Esri Technical Support so we can keep improving Site Scan, the Reality Engine, and your workflows end-to-end.
