Discriminator identifies fakes much easier than reals

Hi I am working on a GAN that generates videos. I have 4 discriminators checking 4 different things in videos. The problem is the accuracy of the discriminators for fakes are much higher than reals. So it can easily identify the fakes but not the reals. Will it create a problem? Also how can I solve it? I searched about this problem but couldn’t find anything.

Thank you