In this guide
Diagnose normal-map seams by comparing the same asset under controlled conditions, with and without the normal map, then checking normals, tangents, UVs and importer interpretation. Do not start by repainting the texture or adding geometry everywhere. First determine which signal introduced the visible break.
Keep the source bake, exported model, texture files and receiving settings together. A screenshot of the final material hides too many variables to explain a source-to-engine mismatch.
Establish a neutral baseline
Use a simple material and a controlled light. Inspect the silhouette and shading without the normal map. If the defect is already present, investigate geometry, vertex normals and transforms before blaming the texture.
Then enable the normal map without changing the camera or lighting. Note whether the problem follows a UV seam, a hard edge, a mirrored region or every face. These patterns narrow the next check.
| Pattern | Compare | Avoid changing at the same time |
|---|---|---|
| Seam exists without the map | Mesh normals and edge treatment | Texture channels |
| Seam appears only with the map | Tangent basis and normal interpretation | Lighting and exposure |
| Mirrored region behaves differently | UV orientation and bake setup | Whole-model topology |
| Relief appears inverted | Channel convention and importer type | Material color |
Unity’s Model import settings document import and calculation controls for normals and tangents.1 Record whether the project uses supplied values or recomputes them; a source bake and receiving tangent basis must be evaluated together.
Check the texture as data
Confirm the file is imported as the intended normal-map type and reaches the expected shader input. Keep packed scalar channels separate from normal-vector data. Do not rely on a filename to establish what the receiver does with the image.
If a channel convention needs changing, make one documented change and repeat the same comparison. Do not flip channels, recalculate normals and change smoothing in one attempt. That may produce a better picture without a reproducible fix.
Preserve the bake-to-low-poly relationship
Ask which low-poly revision was baked. Later triangulation, changed vertex normals or altered UVs can invalidate the relationship. Compare the actual exported receiver mesh rather than only the authoring application’s pre-export object.
If rebaking is required, identify the source correction and create a new revision. Retain the previous bake as evidence of the problem. For a broader material package review, use the PBR handoff guide.
Close with an engine-side retest
Repeat the initial view and inspect more than the originally visible seam. Rotate the light or asset to reveal neighboring surfaces, then check the intended gameplay distance. A detail that disappears only because the camera moved farther away has not necessarily been repaired.
The normal-map diagnosis worksheet records one-variable tests, observed changes and the final export revision. Keep it with the missing texture checklist, since incorrect material assignment can resemble a shading problem. The Rendora game entry connects the supplied-source package and validation evidence; target-engine shading remains part of receiving review.
Shading questions
Should the engine always calculate tangents?
Use the pipeline agreed with the baker and importer. Consistency and an actual receiver test matter more than a blanket setting.
Can increasing texture size remove the seam?
Only if insufficient texture detail is the relevant cause. It does not resolve a wrong basis, channel interpretation or mismatched source mesh.
References
- Unity, Model import settings, checked October 26, 2026.
About this guide
Rendora publishes practical guides for teams working with configurable 3D assets and traceable production handoffs. Updated October 26, 2026.
