Use HDA_Cowgirl position in your positive prompt to generate images.
Remember to put the .pt file in the "embeddings" folder
If you enjoy my content, please consider supporting it ☕☕☕ . Thank you.
Description
FAQ
Comments (6)
Thanks!
Can you share any tips or experiences on TI training for concepts? I tried it a bit but not very successful so far.
Keeps on getting girls with cowgirl hats, any suggestion?
Use "hat" as a negative prompt. Increase the weight if needed ie. (hat:1.4), or just change the file name of the embedding. Unlike LoRAs, embeddings use the file name as the trigger. Changing the name to something unique that SD can't confuse eliminates this problem...something like "SDA_CGPv2" (w/o the quotes). It's something that both LoRA and embedding trainers tend to forget.
Very strange, it has never happened to me and I have made several attempts
@Void91 It may depend on the model being used. I've encountered it a couple of times using Hassaku but not on AbyssOrange (so far). It's no big deal though. Renaming the file took care of the problem. Great work btw!