Advice Needed: Auto Mesh for Tiny Eye ArtMeshes on Full-Canvas PSD Layers - Cubism 5.3

Yandere

Hello Live2D community,

I am preparing a character model for eye-blink rigging in Live2D Cubism Editor 5.3.04 FREE. I would appreciate advice on the best way to generate reliable meshes for very small eye ArtMeshes imported from a 1024×1536 PSD.

Setup

The PSD layers retain full-canvas rectangles, but their visible alpha content is very small:

LEFT_EYE_SCLERA_BASE: approximately 26×14 px

LEFT_EYE_IRIS_PUPIL: approximately 27×9 px

LEFT_EYE_HIGHLIGHT: approximately 6×2 px

LEFT_EYE_LASH_LOWER: approximately 33×10 px, sparse

LEFT_EYE_LASH_UPPER: approximately 44×18 px

The character imports into Cubism with the correct overall composition. I am testing on a separate, verified project copy before adding any blink parameters or keyforms.

What I observed

For LEFT_EYE_SCLERA_BASE, the initial imported mesh was a very large minimal quad. After applying Auto Mesh with the Standard preset and alpha threshold 10, I obtained a localized mesh, but I am not yet confident that its boundary is appropriate for deformation.

When trying the Deformation (Heavy) preset, the mesh appeared much more localized around the eye. However, in a previous test, the generated triangular region appeared not to enclose all of the visible sclera artwork, so I reverted that trial.

For LEFT_EYE_IRIS_PUPIL, Standard appears to produce a mesh spanning a much larger region around the face, while Heavy appears more localized around the eye. I have not yet established the best settings for this ArtMesh.

Questions

For a tiny opaque region on a full-canvas PSD layer, what is the recommended way to choose between Standard and Deformation (Heavy)?

If Heavy creates a localized mesh but appears to leave some visible artwork outside the boundary, what should be checked before using the generated mesh?

Is there a reliable way to determine whether low-alpha pixels elsewhere in the layer are influencing mesh generation?

What objective checks should be performed on a sclera and iris mesh before proceeding to eye-blink keyforms?

I have not created blink keyforms yet. I want to finalize the relevant meshes first and avoid modifying the source PSD or disturbing the correctly composed model.

I can provide cropped screenshots showing the ArtMesh, mesh overlay, and generator settings. I would prefer not to share the complete proprietary PSD or Cubism project.

Thank you for any guidance.