I'm really kind of impressed what's possible now. I haven't focused on any bondage themend stuff for a while because it still was all so awkward and uncanny. It's still not perfect, but the first image of Batgirl getting herself in trouble is the unedited version of a prompt.
Pretty decent right out of the gate.
No LoRa or ControlNet features were used, neither for the position, the hogtie ropes nor the ballgag.
Just plain prompting with the PonyXL checkpoint and some style LoRas and a detail enhancer.
Even the mirror surface of the floor looks almost accurate, the biggest glitch being the face area
AI image generation has huge trouble with upside down faces - as do we humans (see the Thatcher Effect), so who are we to blame it.
But with a little quick and dirty editing, the second image looks pretty aight to me.
Pretty decent right out of the gate.
No LoRa or ControlNet features were used, neither for the position, the hogtie ropes nor the ballgag.
Just plain prompting with the PonyXL checkpoint and some style LoRas and a detail enhancer.
Even the mirror surface of the floor looks almost accurate, the biggest glitch being the face area
AI image generation has huge trouble with upside down faces - as do we humans (see the Thatcher Effect), so who are we to blame it.
But with a little quick and dirty editing, the second image looks pretty aight to me.
(This post was last modified: 10 May 2024, 12:55 by Bound Whore.)