The Screenshot Is Not the Asset

Asset handoffs · By W.T. Roosevelt

Cavi's example is a green utility case: painted panels, dark hardware, a handle and a lid. It is simple enough that a material change has somewhere to be seen. It also makes a useful demand of the tools around it: give the next person something they can inspect and change.

A picture answers what the case looks like. A working handoff also answers which material produced that surface, what was exported, and how another person can make a different choice. A sufficiently confident filename is not a handoff.

The example uses Cavi's Substance MCP with Adobe Substance 3D Painter. Two uniform-fill states keep the painted material's color and metallic value the same while changing roughness from 0.18 to 0.68. The point is a small art-direction decision: how should this particular painted case catch the light?

That question belongs to a person. The useful automation is the work around the decision: applying the stated values, exporting the maps, and making the result inspectable. Calling a material tool does not decide whether the prop belongs in a workshop, a showroom or the back of a truck.

There are two checks worth separating. First, did the exported data contain the intended change? The Painter run produced 16-bit PNG maps. The roughness pixels matched each state's requested value within the declared quantization tolerance. Between the two exports, the intended roughness map changed and the other eleven export files remained identical.

That is more useful than a screenshot of a parameter field. It checks what left the application. It is still a result for these files and settings; choosing a 16-bit container does not recover detail that was never present in a source.

Second, did the editable work survive? The saved Painter project copy was reopened, its fill layers inspected, and the second state exported again. All twelve files from that re-export matched their earlier counterparts exactly. The project could do more than sit attractively in a folder.

The case preview uses the actual Painter-exported maps in Blender. Cavi's Blender MCP supplied material binding and inspection work there. The painted panels' source-UV check recorded no overlaps or degenerate triangles and met the example's two-pixel padding target. That target is a choice for this check, not a universal prescription for texture resolution or viewing distance.

An image, a map check and a source-UV report answer different questions. For an engine handoff, the imported asset needs its own check. A correct-looking Blender preview does not tell us how another renderer interprets the files, whether a prefab keeps its references, or what the asset costs at runtime.

The useful public package is the sample mesh, exported maps, plain settings and short recreation steps. Those inputs let someone build an editable project in their own licensed Painter installation. They also make a disagreement productive: change a value, point to a material assignment, or show exactly where the result differs.

That is a better conversation than asking whether a render is impressive. Someone can prefer the rougher finish and explain why. Someone else can reproduce the task and find a missing instruction. Both leave the next version better informed.

Start with the Substance MCP example and setup, then inspect the Blender workflow where it is useful. Make one small change and check what crosses the boundary. The screenshot can introduce the asset; the files and decisions give the next person somewhere to begin.