Ring_Yiffy37_V4
Strength of 0.6 or 0.7. I used ADetailer to add detail to faces when they were smaller in the images. I trained this one on a new image set and didn't use captions, that seemed to help keep the art style and more of the image layout intact from the model.
ring_yiffy35_v2
Strength of 0.6 or 0.7. I used a smaller file size on the lora and it seems to retain more of the original prompt art style in the final images. Use ring_gag as trigger, do not use open mouth as that wasn't part of the captions and can make the mouth look odd. Can also add drool, tongue out or try large mouth, but they may negatively affect the images.
ring_yiffy3_v2:
A Strength of 0.5 or 0.6 seemed to work sell. It can be very sensitive to negative prompt, I usually didn't use a negative because it tended to make the mouth softer and lower detailed. Use ring_gag as trigger, do not use open mouth as that wasn't part of the captions and can make the mouth look odd. Can also add drool, tongue out or try large mouth, but large mouth can produce some odd results. Used a larger image set. Trained on Yiffymix 3.
ring_yiffy3_v1:
Same as previous, but trained on Yiffymix 3
ring_yiffy21_v1:
A strength of 0.6 or 0.7 seemed to work best. Use ring_gag as trigger, a few other things like drool, saliva, tongue out can affect the results as well. Trained on Yiffymix 21
Description
Trained on Yiffymix3. 0.6 to 0.8 strength. Using ring_gag seemed to get best results, a few other things like drool, saliva, tongue out can affect the results as well
FAQ
Comments (4)
Wish I could get my ring gags looking as clean as yours. Any more info on your generation process?
The images I uploaded have the generation info in them. Anything in particular you were wondering about?
@canngramps Yes. Did you do any Hires upscaling or inpainting on them? Anything beyond the provided generation info. Sorry, Im still pretty new to this.
I use the stable-diffusion-webui by AUTOMATIC1111. I did not do any inpainting. I set the resolution to 512x768 and used the hires fix to generate higher resolutions. I'm not familiar with any other UIs, but in the one I use, you can import the image along with all of the generation information, including the resolution settings, into the UI.