Question on motion input in SMPL form

[OrigamiStationery]

Thank you for your marvelos work. It really helps a lot. But I need instructions about how to use SMPL-formed motion sequences as model input to generate motion descriptions. Desperate for your help.
btw express my gratitude again.

[LinghaoChan]

@OrigamiStationery Thanks for your interest.

The motion format is the humanml format. The motion embeddings are VQ latents of the T2M-GPT.

[vkgo]

@OrigamiStationery Thanks for your interest.

The motion format is the humanml format. The motion embeddings are VQ latents of the T2M-GPT.

Hi Linghao,

I'm attempting to use motion data as input for MotionLLM, but I've encountered a problem. The provided checkpoint download link in the README appears to be missing the checkpoint for the Motion Translator component.

Could you please provide a link to the correct Motion Translator checkpoint, or point me to where it can be found?

Thank you for your assistance.

[LinghaoChan]

@OrigamiStationery Thanks for your interest.
The motion format is the humanml format. The motion embeddings are VQ latents of the T2M-GPT.

Hi Linghao,

I'm attempting to use motion data as input for MotionLLM, but I've encountered a problem. The provided checkpoint download link in the README appears to be missing the checkpoint for the Motion Translator component.

Could you please provide a link to the correct Motion Translator checkpoint, or point me to where it can be found?

Thank you for your assistance.

@vkgo I will publish it soon.