convert2image

Image tools for 3D modelling and texturing

Textures, maps and reference images are just image files, and a lot of 3D prep is image housekeeping. These tools handle the common chores without opening an editor.

Texture sizes

Game engines and real-time renderers work best with square, power-of-two textures: 512, 1024, 2048, 4096. Non-power-of-two sizes can break mipmapping or waste memory once the engine pads them.

  • Make image squarePad or centre-crop a photo texture to a square before scaling it.
  • Resize imageScale a batch of textures to exactly 1024 or 2048. Untick “Don’t enlarge” if you need to scale up.
  • Split image into gridCut a large scan into tiles, or break up an atlas you were handed.
  • PNG to TGAExport 32-bit TGA with alpha, with sides snapped to powers of two.
  • DDS to PNGOpen DXT1, DXT3 and DXT5 textures from games for editing.
  • PNG to DDSWrite uncompressed DDS for engines and mods that load it.

Greyscale and data maps

Roughness, metalness, height, ambient occlusion and masks are single-channel data dressed up as images.

  • Black and whiteTurn a photo into a greyscale height or roughness starting point, or a threshold mask.
  • Invert coloursConvert a glossiness map to roughness (they are inverses), or flip a mask.
  • TIFF to PNGBring scanned or exported TIFF textures into a lighter lossless format.
  • HDR to PNGPreview HDRI environment maps with a tone-mapped PNG and an exposure slider.

Atlases and reference sheets

Combining images keeps draw calls down in engines and keeps references together while modelling.

  • Combine imagesLay out front, side and back reference photos in a single image for your modelling viewport.
  • Crop imageTrim reference photos to the object, with exact pixel boxes for matching views.
  • HEIC to PNGReference photos from an iPhone, as lossless PNGs your 3D software can load.

Colour reference

Matching real materials starts with reading their colour accurately.

Formats for textures

Use lossless formats for anything that is data. Normal maps store surface direction in their colour channels, and JPG compression shifts those values, producing visible bumps and seams in lighting. Height, roughness and mask maps suffer similarly. PNG is the safe default; TIFF and EXR are common in film and offline rendering.

Colour textures can be lossy for delivery. Albedo or base colour textures tolerate high-quality JPG in a pinch, though engines usually recompress textures into their own GPU formats at build time anyway, so there is little gain in feeding them lossy sources.

Mind colour space. Base colour textures are sRGB. Data maps such as roughness, metalness and normals should be read as linear (non-colour) data in your 3D software. The image files are the same; the setting is on the texture node or import options.

Power of two, and square where you can. Mipmapping, compression and streaming all behave best with 512, 1024, 2048 and 4096-pixel textures.

Questions people ask

Why do my normal maps show artefacts?

Usually JPG compression, or the map being read as sRGB instead of linear data. Keep normal maps as PNG and set them to non-colour in your renderer.

How do I convert a glossiness map to roughness?

Invert it. Roughness is one minus glossiness, so the invert tool does exactly that.

What texture size should I use?

The smallest power of two that holds up at your closest camera distance. 2048 is a common default for hero assets, 1024 or 512 for props.