Spaces:

hysts

/

Text2Human

like 73

Running

Hi, I found a bug in the web demo of Text2Human.

The bug is about the inconsistency between the selected pose image and the label image/result image.

The issue is caused when multiple users are trying this demo. For example, User A chooses Pose 1 and User B chooses Pose 2 before User A generates the label image, then User A clicks to generate the label image, and the returned label image is consistent with Pose 2 (chosen by User B) rather than Pose 1.

Could you figure out why this happens? Thanks for your patience!