Describe the surface before making a map
Separate the visible surfaces: exposed metal, painted panels, plastic, printed marks, glass and coated parts. Describe what is known and what still needs confirmation. A metal component underneath opaque paint does not mean the visible paint should be assigned a fully metallic response.
For a technology enclosure or premium consumer product, ask for labelled finish references and consistent views of each region. For an industrial assembly, identify which revision and components the references cover. Photographs show light, exposure and environment as well as the material; they are not a roughness measurement.
Establish a controlled roughness check
Make three opaque grayscale test patches with values 0, 128 and 255. In a normalized eight-bit data input, these correspond to 0, approximately 0.502 and 1. Keep the material factor at 1 while checking that the map reaches the intended slot. This verifies the input path; it does not assign a measured physical finish.
The roughness tool currently has an English interface. Use it to prepare an explicit map, keep camera and lighting fixed, and change one variable at a time. A mid-grey map is not a promise of “50% gloss”. Glossiness-to-roughness inversion is only appropriate when the source and destination definitions agree; a numeric inversion alone is not a calibrated material conversion.
Paint metalness intentionally
Use the metalness editor to mark the intended regions. For a simple separated-surface test, black 0 represents the non-metal region and white 255 the metal region. Keep edge treatment deliberate; an intermediate edge pixel can be a blend, not evidence of a different alloy.
Try the original sample, then clear or edit the mask and inspect the downloadable PNG. The tool does not identify metal from photographic brightness. Save the mask with its material role and apply it explicitly to the correct part or shared material.
Verify channels and distinguish surface detail
A standalone grayscale mask is not an ORM image. In the supported glTF metallic-roughness layout, roughness belongs in G and metalness in B. In the channel packer, a controlled pixel with R=255, G=128, B=0 and A=255 gives a mid-value roughness channel and a zero metalness channel. Split G and B back out to check those values.
Keep source and output dimensions equal for this check and avoid colour correction or lossy JPEG export. The packer preserves numeric channels rather than interpreting a photograph. Its English interface remains labelled as such.
Height describes a separate kind of detail. The height tool’s seeded sample is repeatable, and a flat height input produces a flat normal in the tested connected workflow. That does not measure product relief. In the renderer handoff, height becomes normal detail; the mesh is not displaced.
The grayscale test file contains three pixels, left to right: 0, 128 and 255. The ORM test contains one pixel: R=255, G=128, B=0, A=255. These tiny original files check numeric channels, not UV coverage or visual finish.
Review the finish as part of the product
Keep the same lighting and camera for the first comparison, then inspect the surface under more than one useful lighting direction. Check transitions between materials and whether all parts sharing a material should receive the change. A correct pixel value can still sit on the wrong UV region.
For a production handoff, include the native model, material-region labels, approved references, map roles, channel packing, dimensions and unresolved finish decisions. Explain what the product film or image should communicate. That gives the studio a basis for look development without presenting generated masks as measured materials.
