This LoRA aims to help with shootings in the rain, be it natural or studio rain. The model learned how wet clothing, hair and skin looks and how the rain is present in the image.

However, flux is pretty resilient when it comes to hair from pre-trained characters; that means if you try to use this LoRA together with a character LoRA that had NO "wet hair" in the dataset, you will have a pretty hard time to get the hair wet. I don't know why this is or if I can fix that in a future version but right now it is how it is. Maybe the dataset needs explicit "dry hair" mentioned so that flux understands what to change.

You really have to test the weight for different pre-trained characters; it depends on how well they are trained and how stable the character is. I usually use my character LoRAs at 0.95. Some characters need a maximum weight of 0.4 or the faces will begin to drift. Other LoRAs can even take 1:1 weight. Some need something in between, for example 0.7 - I had different results with different models and there is no single weight that applies to everything.

Use with my "Controlled Sensuality" LoRA to get even more dramatic lighting.

Would love to see what you create with it.

The LoRA was trained on the default flux1-dev model for maximum compatibility.

Dataset and training data:

• 80 images

• 10 steps per image

• 10 epochs

• bucks with 1152px max size, no auto scaling

• cosine scheduler with 0.2 warmup and 0.7 decay

• Captioning was done with OpenAI

• alpha == dim: 32

As always: I did not polish any of the images in the showcase, they all came out like this. I did 2 images per prompt and picked the better looking one.