A lora for much darker and detailed skin tones typical of middle eastern countries and parts of asia. Can also be used at low weighting to create tanned caucasian characters.

Recommended weights: Trigger Words:

Eastern characters: lora weight: 0.5-1.0 - 'dark skin'

Caucasian characters: lora weight: 0.1-0.3 - '(dark skin:0.1)'

Tip: placing the trigger words closer to the start of the prompt (or the very start) will lead to a stronger influence from the lora.

Lora was created with photorealism in mind, using only realistic examples in training. From testing however the lora does also work for more artistic checkpoint models!

Lora was trained on ~500 images to depict skin tones typical of countries such as India, Pakistan and Thailand. This lora is not trained for black/african american characters as I wanted to specifically wanted to target eastern ethnicities for this lora (expect a lora for this in the future though!). The lora was specifically given non-airbrushed training images to avoid blurring/smoothing and retain more skin details.

lora is also trained for ???? models.

Notes:

• By the nature of the training data, the lora will begin to bias towards eastern/black faces as the weighting is increased, can be mitigated with prompts such as 'white', 'caucasian', 'blue eyes'.

• If lora is giving you unwanted ??????, add prompts like 'fully clothed'. In general just prompting for a particular outfit will work.

• For darkest tones, avoid prompting specific ethnicities/nationalities.