KitForma field notes

Choose GLB texture sizes deliberately

Balance image payload against material detail and data-map meaning.

Identify the images that carry the cost

A GLB may contain a small mesh and comparatively large embedded images. The inventory lists each PNG or JPEG image’s dimensions and encoded bytes. Source limits bound both encoded file size and decoded pixel counts, because a small compressed image can still need substantial memory after decoding.

The default keeps the original image bytes. Choosing a maximum dimension resizes only images that exceed that bound, retaining aspect ratio. It does not upscale smaller images or claim to improve their resolution. A new encoded size can be larger for an unusual source, so the actual output size remains the relevant result.

Remember that some textures are data

Color images, normal maps and packed metallic/roughness or occlusion maps have different meanings. The browser resampling path changes pixels and can alter fine detail or normal-map vectors. This desk labels resizing as lossy rather than treating it as a reversible packaging change. It does not guarantee a physically identical material after resampling.

PNG output retains transparency while JPEG is re-encoded with a fixed quality setting. Examine transparent edges, text, hard surface detail and material seams at the expected viewing distance. If a data map needs specialized filtering or normal renormalization, preserve its original bytes and prepare it in a dedicated material-authoring tool.

Use previews as checks, not certificates

The optional orbit preview uses approximate studio lighting and a default scene. It can expose obvious missing textures, geometry issues and animation behavior, but it cannot reproduce every destination’s environment map, tone mapping or post-processing. A preview that looks acceptable here may still need adjustment in the production viewer.

Build separate delivery variants when different destinations have different budgets. Keep each variant’s audit so the texture bounds and animation choices remain clear. Compare the result on a constrained target device as well as a desktop; file-size reduction alone does not measure GPU memory, render cost or total page performance.

The evidence behind the check

Sources & assumptions

Formats, algorithms and supported features are documented below. Check the stated scope and inspect your output before using it.

Reviewed 2026-10-19. Read our methodology or report an issue.

Essential local storage remembers this choice for up to 180 days. Your choice applies to this subdomain. Global Privacy Control and Do Not Track override optional measurement.

Advertising has a separate choice. Use Advertising preferences in the footer to allow or reject the optional guide-page banner.