I tried taking a crack at one of those ThingPunkAI thingies, but as an embedding.
This one makes things magical girly.
Description
FAQ
Comments (10)
Not bad! If you don't mind me asking, why a textual inversion?
That Dreambooth menu was kinda confusing so I gave up. Textual Inversion training was way more straight forward. Also textual inversions are slightly more convenient to use, generate faster, and have a negligible file size...
It's still mostly the giving up, though. š
@KinomiĀ if you get the chance, LoRAs can be made quicker with adjustable file sizes. I still appreciate the work that you've done to adapt what the community has learned via LoRAs into a TI!
Textual inversions are the best! They are so much more flexible when stacking than a LORA. Please continue making them we need more!!!
After a quick experiment I can conclude that while you do need to weight it (I used 1.6 strength), it works great. Results attached below :)
Looks good!
Question: Why a TI instead of a LoRA?
(glancing around while trapped inside a cupcake)
"...I see you're a fan of pink."
A nice crack indeed!! Pleasantly surprised~ š¤š
I cannot get this to work. Does it have to be a certain model? I have it weighted up to 1.8 and nothing. I want this to work so bad!