Photo montages win work: a client understands a design instantly when it stands in the real street instead of floating on a blank background. The hard part has always been matching the 3D camera to the real photograph. BlenderGIS (v2.2.15, GPL-3.0, by domlysz) automates the position half of that problem from your phone’s GPS data — you handle the rest by hand, and it is faster than you think.

What the EXIF data gives you — and what it does not

A geotagged photo’s EXIF records where the camera was: GPS latitude, longitude, and altitude, plus usually the focal length. What it does not reliably record is which direction the camera was pointing. Phones sometimes store a compass heading, but it is rarely precise enough to composite against.

So BlenderGIS gets you most of the setup — camera position and lens — and you finish the last part by hand: rotating the camera until the 3D scene lines up with the photo. Knowing this in advance prevents the most common frustration with this workflow, which is expecting the import to produce a perfectly matched camera on its own.

One prerequisite, and it is non-negotiable: your phone must have location tagging enabled before you shoot. A photo without GPS coordinates gives the EXIF operator nothing to work with.

Step 1: Photograph the site properly

Shoot at roughly eye level from the viewpoints your audience will care about — the street frontage, the main approach, the view from neighboring buildings. Keep the phone level; extreme tilt makes alignment harder. Avoid heavy digital zoom, which narrows the field of view and magnifies every alignment error. Include a few recognizable landmarks — a church spire, a distinctive building corner, a street intersection — because you will use them as alignment references later.

Take more photos than you think you need. You will composite into one or two hero shots, but alternates save you a return trip to the site.

Step 2: Build the site in 3D with BlenderGIS

Back at your desk, build a georeferenced 3D scene of the site: terrain from DEM data plus current basemap imagery, positioned at its true geographic coordinates. If architectural visualization is new to you, start with our guide to BlenderGIS for architectural visualization — this article is the photo-matching specialization of that workflow.

Then run the EXIF camera operator on your chosen site photo. BlenderGIS reads the GPS coordinates from the photo and places a camera at that exact position in your 3D scene, with the focal length taken from the EXIF data.

Step 3: Align the camera to the photograph

Load the site photo as the camera’s background image and look through the camera. Now rotate the camera — position stays fixed, because the GPS already solved that — until 3D landmarks overlay their photographic counterparts: the church spire, rooflines, street corners. Nudge the focal length slightly if the EXIF value feels off.

This is the craft step, and it usually takes 10–20 minutes per view. Small misalignments are visible immediately: trust your eyes over the numbers.

Step 4: Composite your design into the real photo

With the camera matched, model or import your proposed building into the scene, light it to match the photo — time of day matters enormously, so shoot and render with the same sun direction — and render with transparency. Composite the render over the site photograph in Blender’s compositor or an image editor.

If your site photos are noisy, slightly soft, or shot in bad light, clean them before compositing. Denoising tools like Topaz Photo AI can rescue a usable site photo that would otherwise force a return trip.

The result is a photo-grade visualization: your design standing in the real street, ready for client presentations, competition boards, or planning submissions. For early massing studies where photorealism would oversell an unfinished design, consider rendering the massing in the clean white porcelain style instead — it communicates form without implying finished materials.

Honest limits of this workflow

Three things to keep straight. First, GPS accuracy: phone GPS is typically good to a few meters — fine for eye-level views, noticeable if you are matching across a wide plaza. Second, moving elements: people, cars, and scaffolding in the photo will not match your render; plan to mask them or accept them. Third, this is a visualization technique, not a survey — it convinces clients, it does not replace measured drawings.

Frequently Asked Questions

Does BlenderGIS read GPS data from photos automatically?

BlenderGIS includes an operator (add_camera_exif.py) that reads the EXIF GPS coordinates from a geotagged photo and creates a Blender camera at that position. You still align the camera direction manually.

Does EXIF data record which way the camera was pointing?

No, not reliably. EXIF stores the camera position (GPS) and usually the focal length, but direction must be set by hand by aligning the camera to landmarks in the photo.

What do I need to enable on my phone before a site visit?

Turn on location tagging (geotagging) in your camera settings so each photo stores GPS coordinates. Without it, the EXIF operator has no position to read.

What is photo montage in architecture?

Photo montage is the professional practice of compositing a rendered design proposal into a real site photograph, showing clients exactly how a building will look in its actual context.

Can I use photo montages for bids and client presentations?

Yes. Matched photo montages are a standard presentation technique for competitions, bids, and client reviews. They are visualizations, not measured surveys, and should be presented as such.

My site photos are noisy or blurry. Can I still use them?

Often yes. Denoise and sharpen the photo first – tools like Topaz Photo AI can clean up a noisy site photo enough for compositing – and avoid heavy digital zoom when shooting.