Vivi版本v1.0 (ID: 765914)

A LoRA of an unnamed student from Indonesia. Feel free to tinker with the weight and all that, my settings can be found in the featured images.

Some note :

• By default, the image will show inconsistent body feature and hair, since the dataset consists of her wearing covered outfits (hijab, abaya, etc).

• I used epiCRealism checkpoint when testing these, so the quality of image will decrease if you use too many prompts (just my personal opinion, you totally can circumvent these with img2img/inpaint). I highly recommend using ControlNet OpenPose with 0.75 weight for a specific pose, or 0,5 if you use Depth (since Depth will mess-up the facial/body features)

• For those wondering about 'extremely detailed face' prompt, yeah I know it's not actually necessary in the latest epiCRealism checkpoint. Just want to milk out every pixel for those gorgeous skin texture lol

描述:

训练词语: vivi