Training V-JEPA 2 on images

[afilt]

Hello,
Fantastic work, congratulations.
Would it make sense at all to train V-JEPA on images only ? If not, which JEPA architecture would you advise ?

Thanks a lot.