Very slow, any ways to speed up for a local 4090?

[SoftologyPro]

Using the gold dragon head example, on a 4090 24GB it takes (this is from clicking the Generate button until the mesh appears in the UI)...
Generate Shape Only - 4m26s
Generate Shape And Texture - 8m0s
Using text prompt "a 3D model of a rose on a white background" - Generate Shape Only - 10m7s
Using text prompt "a 3D model of a rose on a white background" - Generate Shape And Texture - 15m44s

GPU is maxed out at 100% usage and 24 GB VRAM is also maxed out.

Any enhancements possible that can reduce VRAM usage and processing time?

[YanWenKun]

mmgp may do the trick:
https://github.com/deepbeepmeep/Hunyuan3D-2GP

don't forget to set profile level:
https://github.com/deepbeepmeep/mmgp?tab=readme-ov-file#usage

--profile 1

[SoftologyPro]

I did end up using your version
https://github.com/YanWenKun/Comfy3D-WinPortable/releases/tag/r8-hunyuan3d2
which works fine for lower VRAM and speed. Thanks!

This issue was more to see if there is a reason the original code is so slow. Even on a 24 GB 4090.