Does TRELLIS.2 support 3D-to-3D generation / editing?

[xzr52]

Thanks for releasing TRELLIS.2 — it’s been very helpful.

I’m currently using the Image-to-3D pipeline, and I’d like to ask whether TRELLIS.2 supports 3D-to-3D generation/editing (e.g., taking an existing 3D asset such as mesh/voxel/SDF as input and producing a modified/retargeted 3D output).

Any roadmap/plan for 3D→3D support would be greatly appreciated.

[mr-lab]

logically it can ! if you manage to convert your 3d model information that the model can understand .
i might look into it
since the biggest draw back with this project is to generate the whole subject from a single image , maybe after generating the first 3d model from a single image you select a specific region on the 3d model and refine it , like hands
so same logic goes for editing an existing 3d model , it will take some time to think about this thanks for the idea <3

[francisengelmann]

You might want to have a look at SpaceControl (also based on Trellis) which does something along those lines:
https://spacecontrol3d.github.io/