Free, in your browser, no sign-up
Roughness Map Generator
Get a starting roughness map from any texture, together with matching normal, height and ambient occlusion maps. Download one PNG or all four in a ZIP.
PNG, JPG or WebP. Large images are scaled to 2048 px. Transparent sprites keep their outline.
Free maps are made on your device. Nothing is uploaded unless you ask for an AI set.
Roughness
AI texture set
An AI model reads the material in your photo and returns base colour, normal, roughness, metalness and height. Preview and export it like the free maps.
Sign in to get a free AI setExport for
ZIP with BaseColor, Normal (DirectX), ORM packed, Height and a setup note.
What a roughness map does
In a PBR material, roughness says how blurry reflections are. Black means smooth and glossy, white means matte. A good roughness map is what makes wet stone look different from dry stone.
This tool estimates roughness from the height of the surface: crevices and dark gaps come out rough, raised and bright parts a little smoother. It never goes fully glossy, because scanned textures rarely are.
When to adjust it by hand
Treat the result as a base. For metal, polished wood or anything wet, darken the shiny areas in an image editor. If your engine wants glossiness instead of roughness, invert the image.
Need maps from a real photo?
The free maps are worked out from brightness, which suits clean textures. For photos of real surfaces, AI texture sets read the material itself.
Free maps
Free
Runs on your device
- 8 maps, including packed ORM
- Ready-made sets for Unreal, Unity, Godot and Blender
- Worked out from image brightness: best for clean textures and sprites
- Unlimited, no account
AI texture set
From $0.20 a set
1 free set when you sign up
- Base colour, normal, roughness, height and metalness
- Made by PATINA, an AI model built for PBR materials
- Tells metal from non-metal parts, which brightness alone cannot
- Packs of 15, 50 or 150 sets, valid for a year
Questions
Roughness or glossiness?
They are opposites. Unity (Standard, URP and HDRP) uses smoothness, read from the alpha channel of the metallic or mask map, so invert the roughness map and put it in that channel.
Can it tell metal from plastic?
No. From one image it can only guess roughness from the surface shape. Material type needs your judgement or an AI model trained on materials.
Do I need a roughness map?
Without one the whole surface reflects light the same way. Even a simple roughness map makes a texture read much more like a real material.