A lora based on my favourite artist of all time inputwo
if you like his please consider supporting him!
Image info
Model used abyssorange2hard
Training settings?
The most important part of training is getting a good dataset!
Training settings can be found by installing the additional networks extension in auto1111
Extensions -> available -> load from -> additional networks
Restart webui
place lora into stable-diffusion-webui\extensions\sd-webui-additional-networks\models\lora
To view the training settings
Go to additional network tab -> choose model -> training parameters
Description
FAQ
Comments (2)
Have you already shared your thoughts on what a 'good' dataset entails elsewhere and do you mind linking me there? If not, do you mind sharing what qualities you look for when compiling images? I hear all kinds of contradictory things from all sorts of people and thought I would try asking you, seeing that you've trained several nsfw models already and seem to get good results.
Your not the first person to ask me that, so I'll defiantly write up a guide on how to gather a good dataset. As for the short explanation. character Lora, make sure the traits of the character is present in each image, ie the same hair color/bodytype as well as making sure only the character you are training is present in the images (cropping is my go to method). For a style, make sure the style is consistent. one thing that KILLS style loras is not getting images that have the same style. Having a smaller dataset with a more consistent style > big dataset with less consistent style. for concepts making sure each image is relevant ie making sure each image depicts a full blowjob instead of just licking, if your making a blowjob lora. as for my go to place for gathering data