This is my very first LoRa. Please tell me if something needs to be corrected or enhanced.

49 images were used as source. They were generated from "slave dress" concept by Pony (ponyRealism & autismmix): 41 photorealistic + 8 anime style, size = 512*512

I cannot use the trigger "slave dress" in Flux because it generates black women wearing rags.

Made in 1 hour with a 5070 with FluxGym with parameters below:

- Repeat = 8

- Epoch = 5

- LR = 2e-3

- network_dim = 8

- gradient_accumulation_steps = 7

- text_encoder_lr = 3e-4

- network_alpha = 2

The animation was generated by Wan from three pictures from various angles.

Tip :

- The LoRa failed to learn "full body shot". If you want this, you shall talk about her feet or shoes, for example "barefooted" or "high heels".

- Precise the color of the dress, otherwise you'll get grey or brown.