Post #2374296

2026-04-30 17:57 UTC

@nothings@mastodon.gamedev.place Interesting question! I'm not entirely sure :) This would run into problems in UV-degenerate regions but you could solve them with a synthetic orthogonal basis. Ofc none of the techniques I describe further in the post would apply to non-orthogonal bases. I'd guess the industry just gave up on skewed mapping. There's extra complications wrt TBN in that you might need a matrix *inverse* instead of just transpose in certain parts of the pipeline. And vertex averaging muddles it anyway.

Replies (3)

@absurdistcode@mastodon.gamedev.place 2026-04-30 18:04

@zeux@mastodon.gamedev.place @nothings@mastodon.gamedev.place I wonder if two low quality tangents and bitangents could be enough to reconstruct a high quality normal. it doesnt matter if (bi)tangents are a few degrees off, so they could be rotated slightly until a better normal is reconstructed.

Open ##2374298
@nothings@mastodon.gamedev.place 2026-04-30 18:36

@zeux@mastodon.gamedev.place I mean, plausibly, there's no compact encoding of two non-orthogonal vectors (e.g. no unit vector+angle), so you just can't hit the target quality while allowing skewed UV, so skewed UV was discarded. But the thing is, even though it COULD have been a conscious choice, I don't have any evidence that it really a conscious size/quality choice, rather than just "let's do this because it's easier". Which is just kinda weird.

Open ##2374299
@datenwolf@chaos.social 2026-05-01 22:30

@zeux@mastodon.gamedev.place @nothings@mastodon.gamedev.place Fun fact: inverse transpose isn't even the correct generalization: https://terathon.com/blog/transforming-normals.html

Open ##2374309