This model is based on the beautiful HYC. Trigger word is hoacb. Work with all styles that I've tried so far. Always use vae-ft-mse-840000-ema-pruned. You can do eye experiments for changing her eye colors, eg: Deep Blue Eyes, Hazel Green Eyes. Play around with her hair as well, she works on Blonde, Black, Redhead and naturally is a Brunette.
For best results use Heun for sampling at 20-35 (so far tested).
CFG scale so far 9 (I use as default) have been working best.
Description
This model is based on the beautiful HYC. Trigger word is hoacb. Work with all styles that I've tried so far. Always use vae-ft-mse-840000-ema-pruned. You can do eye experiments for changing her eye colors, eg: Deep Blue Eyes, Hazel Green Eyes. Play around with her hair as well, she works on blonde, redhead and naturally is brunette.
For best results use Heun for sampling at 20-35 (so far tested).
CFG scale so far 9 (I use as default) have been working best.
FAQ
Comments (9)
Hello! May I ask for an embedding or lora model? Checkpoints have taken lots of my storage. ;P
tbh. thats a you issue, not the creator's issue
@civitfan01 Loras have a huge benefit of being able to be combined. Checkpoints are great for larger models that really finetune and add a lot of capacity to the base. Loras are great for smaller sets of added information which benefit from combining with other sets. Embeddings are great for known people, styles already present in the datasets, etc.
LORA extract has been added, not perfect but try using both of the LORA models together with the same trigger, it's works better together. CFG scale 3-4 seems to work well.
Would be great to have a safetensor
Thanks for the suggestion will work on it.
Seem there are problems with the fingers ... maybe it requires the bad_prompts in the negative