The 'Doll-Series' is a set of LORA focused on realistic Asian faces, with incredible levels of beauty and aesthetics.
This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage.
The CreativeML OpenRAIL License specifies:
-
Make sure you are aware on the usage instructions of LORA
-
VAE: vae-ft-mse-840000-ema-pruned (For realistic models)
-
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
-
Steps: 20 to 35
-
Clipskip: 1 or 2
-
Upscaler : Latent (bicubic antialiased)
-
CFG Scale : 5 to 9
-
LORA weight for txt2img: anywhere between 0.2 to 0.7 are recommended
-
Denoise strength for img2img: 0.4 to 0.7
The Taiwan variant version, I have decided to discontinue v10, it is still out there somewhere in the internet, you may still find it.
This version is a huge overhaul and remake, instead of building upon v10 or v15, I took small amount of samples from v10 and introduced a new pool of training images.
The China variant version, took awhile despite the requests, will probably make more versions of it in the future.
-
Q: Why can't I produce the same pictures as you?
-
A: Sorry I don't share my prompt, you may check the recommended settings, but you may ask me for advice on Twitter or Pixiv
-
Q: What is the difference of each version upgrade?
-
A: Version upgrade does not mean it will fix hands or legs, it is mainly difference of the face of the LORA, newer versions have wider range of faces.
-
Q: Will you release all your other LORAs?
-
A: Yes, maybe, but I would like to keep certain LORA to be exclusive to fans or supporters in the future through fanbox/ko-fi
-
Q: Do you take requests or commissions on making custom LORAs?
-
A: I might do community voting for requests, if you are somehow interested in a custom/exclusive LORA, you may contact my through Pixiv or Twitter for discussion
