This version is more balanced by character, but it has even more concepts, and that is too many for this number of steps, so it isn't really better and does not support floats well. Balancing was done not by increasing repetitions of small concepts, but by decreasing the loss weight of big concepts, so Zecora, Celestia and Luna got the most steps, even if they didn't have as much influence on each one.

The trigger syntax is like that of the other v2 versions, but with "inflatable" (not "dirlpinf" or "ddptponix").

See the other 2.x versions for more details.

Trained using the Pony Diffusion V6 XL checkpoint, so what is allowed by the licensing terms on this page is further restricted by its license.