Trained on 21 photos, mostly front shots, 768x512, 3150 steps, 1 epoch, NO CAPTIONS.
Rename to ¨xw1ng.safetensors¨ to match prompts.
This one is dicey. You might get an Asterisk-Wing sometimes or extra wings or laser cannons. Nothing you can't fix with Generative Fill or inpainting. I tried several combinations of steps and epochs and this is the best result. I think SD can't just figure out the shapes. Use at your own risk.
Trained on 21 photos, mostly front shots, 768x512, 3150 steps, 1 epoch, NO CAPTIONS.
Rename to ¨xw1ng.safetensors¨ to match prompts.
This one is dicey. You might get an Asterisk-Wing sometimes or extra wings or laser
cannons. Nothing you can't fix with Generative Fill or inpainting.
I tried several combinations of steps and epochs and this is the best result. I think SD can't just figure out the shapes. Use at your own risk.