ORM vs MRA vs RMA Channel Packing

The acronym is not a file-format guarantee

PNG, WebP, TGA, and other image containers do not know that one byte means roughness and another means metallic. The renderer learns that meaning only from a material binding, shader graph, or file-format texture reference.

That is why a channel swap can remain silent: the image decodes correctly, but the shader reads valid numbers for the wrong material property. Preserve a manifest or project convention beside the packed file.

Packing and compression are separate decisions

Packing reduces texture bindings and file count; it does not guarantee lower decoded GPU memory. The runtime compression format, alpha presence, mip chain, dimensions, and platform transcode determine the stored and resident cost.

Review highly detailed masks after compression. Epic notes that channel quality can vary and that an alpha channel can increase memory, so put the most artifact-sensitive data where the target compressor preserves it best and measure the actual build.

Pack the contract, then prove it in the destination

The letters are the channel order: ORM is R=occlusion, G=roughness, B=metallic; MRA is R=metallic, G=roughness, B=AO; RMA is R=roughness, G=metallic, B=AO. Match the destination shader—not the filename.

Step 1: Load the loose scalar maps or ZIP

Add ambient occlusion, roughness, and metallic sources to Texture Channel Packer. Confirm each detected role before applying a preset.

Open Texture Channel Packer

Step 2: Choose the documented destination contract

Select ORM, MRA, RMA, glTF ORM, Unity HDRP Mask Map, or Custom. If the destination has no published contract, mirror the actual material graph and record the choice.

Step 3: Inspect red, green, and blue separately

A composite preview can hide a swap. Solo every channel and compare it with the labeled AO, roughness, or metallic source before export.

Step 4: Export linear PNG plus the manifest

Keep scalar material data out of sRGB conversion. Use the documented ZIP when the mapping, neutral constants, or resampling decisions need to travel with the file.

Step 5: Validate the imported material

Bind the exact channels, rotate the light, and run Material Preflight before the asset ships. A correct-looking RGB thumbnail is not an acceptance test.

Run Material Preflight

Common Pitfalls

• Metal looks like chalk or plastic because the material reads AO or roughness as metallic.
• Cavities brighten because AO polarity was inverted or the packed map was gamma-decoded as color.
• Gloss runs backward because a smoothness source was packed without 255 − value inversion.
• The browser preview looks plausible but the engine is wrong because the importer or shader bindings use another contract.

Are ORM, MRA, and RMA interchangeable?

No. They can contain the same three scalar maps, but those maps occupy different RGB channels. A shader that expects ORM will silently read the wrong values from an MRA or RMA texture unless you rewire or swizzle the channels.

Does Unreal Engine require ORM?

No. Unreal materials can read any channel and connect it to any mask input. ORM is a common export and project convention, including Adobe’s documented Unreal preset, but the material graph remains the final contract.

Is glTF ORM one texture or two?

glTF defines a metallicRoughnessTexture that reads roughness from green and metallic from blue, plus an occlusionTexture that reads red. Both bindings may reference the same image when the UV coordinates match, producing the familiar ORM file.

Should a packed material texture use sRGB?

Normally no. AO, roughness, metallic, smoothness, and masks are scalar data. Import the packed texture as linear or non-color data unless the destination specification explicitly says otherwise.

Do I need separate MRA and RMA tools?

No. Use presets in one general Texture Channel Packer, or define a custom R/G/B mapping. Keep the fixed ORM Texture Packer for the fast legacy ORM job and use PBR Engine Converter when you need a complete target-specific material package.