Combining Photos in RealityCapture

How to Prepare Photos and Build a 3D Model in RealityCapture – Practical Guide

Below is a short, practical guide outlining step-by-step how I prepare photos for creating 3D models using RealityCapture. This is a proven workflow I apply both in drone work and handheld photography. It allows for high-quality model generation without unnecessarily overloading the computer or wasting time.

1. Shooting in RAW Format

All photos – whether taken by drone or handheld camera – are captured in RAW format. This gives full control over exposure, color balance, and detail, which is crucial for later editing and model generation.

2. Developing RAW Images

I use Adobe Lightroom for basic post-processing. The adjustments I typically apply include:

  • gentle sharpening of the image,
  • increasing color saturation,
  • lifting shadows and reducing highlights to balance the photo’s dynamic range.

The editing should remain subtle – the goal is to improve clarity and legibility, not to “beautify” the image.

3. Exporting to JPG Format

Once processed, the images are exported from Lightroom to JPG in full resolution.
Importantly, GPS metadata (including RTK data from drones) stored in RAW files is preserved in the exported JPGs, so no positional data is lost.

4. Importing into RealityCapture

At this point, I import the JPGs into RealityCapture and begin the 3D modeling process.

5. Why I Don’t Use RAW (DNG) Files Directly

While RealityCapture (and tools like Metashape) support importing RAW files directly, I do not recommend doing so.
Processing RAWs directly is computationally heavy and slows down the workflow.
It’s much more efficient to prepare optimized JPGs and work with them instead.

6. Combining Drone and Handheld Photos in One Project

In many real-world scenarios, I combine drone images with handheld camera shots. This is often necessary when drone operation becomes limited – for example, due to time restrictions or the presence of people.
A good example is when I scanned the main gate at Auschwitz: I captured some shots with a drone, then completed the set using a handheld camera once flying was no longer possible.

Here’s how to combine both datasets into one model:

Step 1: Prepare the Photos

Both datasets (drone and handheld) must first be developed from RAW to JPG at full resolution.

Step 2: Import and Align Drone Photos

Start by importing only the drone photos into RealityCapture.
Run Align Images to generate the first model – Model 1, which will be georeferenced (based on RTK data).

Step 3: Disable Drone Photos Temporarily

Once aligned, select all drone images (CTRL+A) and disable them from further processing using CTRL+R.
This prevents them from being considered during the next alignment stage.

Step 4: Import Handheld Photos and Align Model 2

Now import the handheld ground-level photos into the same project.
Run Align Images again – this time RealityCapture will align only the handheld photos and generate Model 2.

Step 5: You Now Have Two Separate Models

At this stage, the project contains:

  • Model 1 – from drone images (currently disabled),
  • Model 2 – from handheld images.

The goal now is to merge both into a single coherent 3D model.

7. Merging Models Using Control Points

Step 1: Re-enable the Drone Images

Re-select all drone images and re-enable them using CTRL+R.

Step 2: Add Shared Control Points

You now need to indicate common reference points visible in both datasets (drone and handheld).
You must define at least 4 shared Control Points, but 5 or more is recommended for greater accuracy. Distribute them evenly across the structure.

How to do it:

  1. Go to the Control Points tab.
  2. Add Control Point 1 and mark it on several drone photos.
  3. RealityCapture will suggest additional matches automatically.
  4. Then switch to handheld photos and mark the same Control Point 1 on the corresponding images.
  5. Repeat the process for Control Points 2, 3, 4, 5, etc.

This is a critical step – it tells RealityCapture how to spatially align both models together.

Step 3: Merge the Models

Once all shared points are in place, run Align Images again.
RealityCapture will generate a new model – Model 3 – that merges both the drone and handheld datasets into one unified scene.

Step 4: Final Steps – Mesh and Texture

From the merged model, proceed with:

  • Mesh generation – using the Normal Detail option,
  • Texturing – using the Texturing option.

The model is now ready for further editing or export.

In Conclusion

I understand this workflow might seem complex at first – I’ve personally spent many hours refining it and learning its nuances.
But once mastered, it becomes a powerful way to create detailed, accurate models even from challenging mixed datasets.

If you run into any issues, feel free to share a downloadable link to the full set of photos (both drone and handheld) of the main gate at Auschwitz.
I’ll be happy to prepare the RealityCapture project for you – with images aligned, control points set, and the merged model ready to import into your own RealityCapture instance for further processing or export.

Scroll to Top