Weird result, when using gltfpack optimized glb

[blanxdev]

I have a strange glb model, that renders with problems in Unity 6.3 when it is compressed via gltfpack.
I think the reason is because of vertically flipped UV in the model.

The only way to get it right in Unity Runtime was by configuring gltfpack to export with floating point texture coordinates and vertically flipped textures.

Which is strange because other tools did not have the problem to display the compressed model correctly.

Original

Old GltfPack

Recent GltfPack

Compressed with floating point UV and flipped texture

[atteneder]

Hi @blanxdev

Could you please provide the following:

• The exact gltfpack versions used
• The gltfpack commands used to convert
• Ideally the glTF model as well. Makes it easier to rule out model differences during reproduction.

The flipped UVs might be caused by broken UV conversion jobs, so having the model would help narrowing down the scope.

I suspect the burger being black is due to Unity's precision limits. gltfpack by default applies mesh quanitzation and scales down the glTF nodes to very small values. Try scaling it up by about factor 10 and see if at some point the colors kick in. The only way to make sure this is not a problem is to provide the -noq option to gltfpack. It'll disable some size optimization, but at least those models now work again.

hth

[blanxdev]

Am Do., 26. März 2026 um 14:26 Uhr schrieb Andreas Atteneder < ***@***.***>:

…

*atteneder* left a comment (atteneder/glTFast#814) <#814 (comment)> Hi @blanxdev <https://github.com/blanxdev> Could you please provide the following: - The exact gltfpack versions used - The gltfpack commands used to convert - Ideally the glTF model as well. Makes it easier to rule out model differences during reproduction. The flipped UVs might be caused by broken UV conversion jobs, so having the model would help narrowing down the scope. I suspect the burger being black is due to Unity's precision limits. gltfpack by default applies mesh quanitzation and scales down the glTF nodes to very small values. Try scaling it up by about factor 10 and see if at some point the colors kick in. The only way to make sure this is not a problem is to provide the -noq option to gltfpack. It'll disable some size optimization, but at least those models now work again. hth — Reply to this email directly, view it on GitHub <#814?email_source=notifications&email_token=AGYXOBN5TVORKWOYVHP52KD4SUVXZA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJTGQ3TOMRXHEYKM4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-4134772790>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGYXOBKTLBRACR433S2ZQLT4SUVXZAVCNFSM6AAAAACWYDORH6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCMZUG43TENZZGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[atteneder]

@blanxdev seems like an inadvertent message to me, was it?

Anyways, looking forward to your response.

[blanxdev]

The message should have contained the glb model

checker.glb.zip

[atteneder]

@blanxdev Thanks,

I can confirm that with gltfpack 0.25 this creates a completely black (or white) burger:

gltfpack -cc -i checker.glb -o checker-cc.gltf

Reason being is Unity cannot handle the precision with such low local scale values:

Increasing the local scale by roughly 5x fixes the shading

Work around this by disabling quantization (which sacrifices some of the compression):

gltfpack -cc -noq -i checker.glb -o checker-cc-noq.gltf

I noticed that updating to gltfpack 1.0 (more precisely git has edbbfba60edfb50b1bf571d078ba53302839e3bf) did not change the results.

To make sure the supposedly flipped textures aren't a result of KTX compression I also tested this:

gltfpack -cc -noq -tc -i checker.glb -o checker-cc-noq-tc.gltf

The texture was applied correctly.

Can you confirm (by using the same version/commands) that this aligns on your end? If not we need to dig deeper where the differences are.

Thanks