Scanned Textures: When Photographing Reality Beats Painting It

8 min read · Last updated July 2026

A camera photographing a mossy rock surface on the left, resolving into a set of PBR texture maps on the right
A real surface, measured — not painted. That fidelity is the whole appeal of a scan, and also its baggage.

Walk through almost any modern AAA environment and a large share of what you are looking at was photographed, not hand-painted. The shift to scanned textures — real surfaces captured into PBR maps — is one of the quieter revolutions in game and film art, and it changed what “realistic” even means. It also comes with a set of trade-offs nobody mentions until your pristine 8K scan refuses to tile. Here is how scanning works, where it shines, and where a generated texture quietly wins.

What a scanned texture actually is

A scan is a measurement of a surface rather than an interpretation of one. Instead of an artist deciding how rough a stone is, a camera records it, and software solves for the maps that reproduce it: base colour with the lighting stripped out, a normal that carries every real bump, a roughness that varies exactly where the stone is worn, and often a height map for the deep detail. Because it is measured, a good scan has a kind of irregularity — the asymmetry of real wear — that hand-authored and even generated materials work hard to imitate.

How a surface gets captured

A ground surface being photographed from multiple angles with a camera on a rig under even lighting, illustrating a photogrammetry capture setup
Many overlapping photos under controlled light. The maps are solved from the pixels; the shadows are removed afterward.

There are two families. Flat scanning photographs a patch of ground or wall under even light with a polarising filter on both the lens and the lights — the polariser kills specular reflection, which is what lets the software separate true colour from shine. Photogrammetry takes dozens of overlapping photos of a rock or a wall from many angles, rebuilds a high-poly mesh from the parallax, and bakes that geometry down into maps. Both end the same way: delighting, the step that removes any baked shadow so the material lights correctly in your scene rather than carrying the sun of the day it was shot.

Scan, generate, or paint — three ways to the same maps

It helps to see scanning as one of three sources, each with an honest cost:

  • Scan — perfect fidelity to a real surface, but only surfaces you can physically shoot, and you inherit the shoot’s lighting and resolution. Heavy files, real gear.
  • Generate — any surface you can describe, seamless by default, produced in seconds; an interpretation, not a measurement. See text to texture AI.
  • Hand-author — total artistic control, and the only real option for stylised or fantastical materials, at the cost of the most time and skill.

Most serious pipelines mix all three: scans for the hero ground and walls, generation for the long tail of surfaces and for anything that cannot be scanned, and hand-authoring for the stylised pieces.

Where scans quietly let you down

The fidelity is real, but so is the baggage. Three things catch people out. First, scans are not seamless — a scan is a specific patch, so its edges do not tile until you make them, usually with a seam editor. Second, lighting can survive delighting — a mediocre delight leaves soft shadows baked into the colour, and they resurface as ghost darkening under your own lights. Third, resolution and storage — an 8K scan set with five maps is a lot of memory for a surface that might be twenty metres from the camera. A scan is not automatically the right call just because it is the most accurate.

Scanning without the rig

A single flat phone photo of a brick wall being converted into a clean set of PBR maps without a studio setup
One flat, evenly lit photo is a lightweight scan. No polariser, no rig — just the surface you walked past.

You do not need a studio to borrow the core idea. A single flat, evenly lit photo of a surface, run through a photo-to-PBR tool, produces the same kind of maps a scan would — perspective corrected, delit, and split into colour, normal, roughness, AO, and height. It is a lighter, faster form of scanning: the fidelity ceiling is lower than a full polarised capture, but for most props and background surfaces the difference never reaches the screen. And because the tool wraps the edges, you skip the seamless problem that a raw scan hands you.

When to scan and when to generate

A rough rule: scan when the exact real surface is the art — a specific cliff for a location, a signature hero material the camera lives on. Generate when you need breadth, speed, seamlessness, or a surface that does not exist to be photographed. And reach for photo-to-PBR when you have a real reference in front of you but no reason to haul out a rig. The three are not rivals so much as different tools for different distances from the camera.

Scan-quality maps from a single photo
Photo-to-PBR turns one flat shot into colour, normal, roughness, AO and height — seamless, no rig. Or describe a surface and generate it. Free, CC0.
Open Studio →

Try CraftPBR

  • Photo-to-PBR — a flat photo becomes a full, seamless PBR material
  • Text-to-PBR — describe any surface, scanned or not, and generate it
  • Free toolsseam editor, normal map, height map
  • Engine export — Unity, Unreal, Blender, Godot, Three.js
  • CC0 license — everything you make is yours

Try CraftPBR free →

Frequently asked questions

What are scanned textures?

Scanned textures are PBR materials captured from a real-world surface instead of painted or generated. A photographer shoots the surface — often under controlled or cross-polarised light and from many angles — and software reconstructs the colour, normal, roughness, and height maps from those photos. The result is a material that carries the exact grain, wear, and irregularity of the real thing.

How are textures scanned?

Two common methods. Flat-surface scanning photographs a patch of ground or wall under even, polarised light to separate colour from reflection, then derives the maps. Photogrammetry takes dozens of overlapping photos of an object from different angles and rebuilds a high-poly mesh plus textures, which are then baked down. Both finish by delighting the images so no baked shadows remain.

Are scanned textures seamless?

Usually not out of the box. A scan captures a specific patch of a surface, so its edges do not tile on their own — you still have to make it seamless afterward by healing the seams or letting a tool wrap the edges. This is one of the biggest surprises for people who assume a high-end scan is ready to tile.

What is the difference between scanned and AI-generated textures?

A scan reproduces one real surface with total fidelity but only the surfaces you can physically photograph, and it inherits the lighting and resolution of the shoot. AI generation invents any surface you describe, seamless by default, but is an interpretation rather than a measurement. Many pipelines use both: scans for hero realism, generation for coverage and for surfaces you cannot scan.

Do I need special equipment to scan textures?

For studio-grade results, yes — even lighting, a polarising filter on the lens and lights, a colour chart, and a tripod. But a single flat, evenly lit phone photo run through a photo-to-PBR tool gets you a usable material without any rig. It is a lighter form of the same idea: turn a photograph of a real surface into PBR maps.