Training IP-Adapter without text prompt

[SantiagoJN]

Hi,

In the original paper it is stated that (page 6, footnote 1)

Note that it is also possible to train the model without text prompt since using image prompt only is informative to guide the
final generation.

What does it refer to exactly? Setting an empty string as text prompt? Fixing some general prompt like "image"?.
Has anyone managed to explicitly train a model that doesn't use text prompt whatsoever?