Disable quantization during CLI mesh optimizing

[bolatuly]

Hi! mesh optimizing uses KHR_mesh_quantization extension which is not supported by application I am using.
Is it possible to disable quantization during gltf-transform cli function meshopt?

To be able to do something like this:

gltf-transform meshopt input.glb output.glb --no-quantization

something like gltfpack tool which has option -noq to disable KHR_mesh_quantization

[donmccurdy]

@bolatuly could you share which flags you're setting on gltfpack, if that's providing results you prefer? If only -noq is specified then gltfpack skips meshopt compression entirely, and just does general optimization on the model. In glTF Transform that would be comparable to:

gltf-transform optimize in.glb out.glb --no-compress --no-instance

If you're also using the -c or -cc flags in gltfpack, then this does the same general optimization, then also applies meshopt compression, which would require your application to support EXT_meshopt_compression. In my experience it's unusual for an application to support EXT_meshopt_compression without also supporting KHR_mesh_quantization; they work best when used together. gltfpack offers the option to apply meshopt compression withou quantization, but glTF Transform currently does not. If that's an issue, I'd be interested to hear more about the use case and/or application if possible!

[bolatuly]

I use both -c and -noq options for gltfpack. Application used by me for model render is Mapbox GL JS, they don't disclose much about the specific extensions they support for glb models. However, when I try to render a model with EXT_meshopt_compression, it works fine. But if I add KHR_mesh_quantization, it renders with some artefacts. I looked through their code and only found references to EXT_meshopt_compression support.

[donmccurdy]

I see, thanks! You're right, they support EXT_meshopt_compression but not KHR_mesh_quantization. It might be worth filing an issue for them about that, it's usually one of the simplest glTF extensions to import, but regardless. I've opened a new feature request about supporting meshopt compression without quantization. I can't promise any timelines for this, but PRs are always welcome.

• Allow writing EXT_meshopt_compression without quantization #1524

In the meantime, Draco compression might be the next best option.

[bolatuly]

thanks for your reply! Originally I used Draco compression, but I liked the decoding speed of meshopt compression, so I started trying it out.

[donmccurdy]

In progress:

• feat(extensions,functions): Allow encoding meshopt-compressed float32 accessors #1526