Tried to make some animorphis slider with leco for xl. This is one of the tests and it seemed to work too well to not throw it here. Not really my forte, but @xmattar .... here is your damn furry slider.
How to use
Gives Human on negative, anthro on positive.
Weight depends on length of prompt and other loras used. Somewhere between -2 and 2 seems to work decently most of the time.
Using same seed gives somewhat same composition on image giving decent "transformation" effect.
Tags:
human, anthro . Seems to work decently by just using both tags, and depending on lora weight. It gives what it gives and takes what it takes.
Personally used both source_furry, source_ anime, on tests, seemed to work.
Tags like species transformation might help. Don't really know furry terminology well so... test.....
Using solo can reduce some random animals, multiple people.
Generally should work with any species.
If you want to keep the "transformation" very similar to original human and want to add tags, i recommend using prompt editing
example: [ :torn clothing:0.25], will add torn clothing at 25% of generation
more on how this works: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#prompt-editing
Description
Initial release
FAQ
Comments (9)
Looks great! Is there any way you could get this to work for regular Stable Diffusion and not XL?
Originally i tried to do this with 1.5. There was just too much bleed for it to work nicely. Never tried it with any furry models tho. Might give it a try if i find time for it.
@PotatCat i know some models can help with decreasing and sometimes rmeoving bleed, counterfiet for example is really good at it, u can add it in combination with negatives and masterpeiece prompts
thats according to my experience atleast
very nice!
i have awoken from my slumber
Can i have your dataset and train settings? Your lora performance is GREAT!
Thank you, this is a ileco trained with https://github.com/hako-mikan/sd-webui-traintrain It uses the knowledge of the model and does not require dataset.
Have not perfected the training settings yet for XL, the basic settings work decently, so i recommend starting from there. (～￣▽￣)～