Style > gpt-image-2

It steers the model towards the style of gpt image 2. Improving details, lighting and texture a lot, though the image might end up looking slightly sloppy. This is meant as an enhancer so play with lora strength to allow more style diversity :)

This is trained without a reg dataset, so the trigger word is not needed, but helps ever so slightly.

V1.1 (when civitai chooses to tag images) is closer to the original images, it adds a lot more texture and details, shift the colors less, and affect the pose / composition a bit less too. It has the downside of affecting the face more than v1.

You might just prefer the effect that V1 give if you want less texture or more punchy colors.

I found some better optimizer settings, and fixed some bug in the custom code I added to the trainer that was most likely holding back all of my anima LoRA. Which explain the large change between the two version when trained with the exact same dataset.

I might train another version since the v1.1 LoRA was still learning when the save used was made, but it started to get bad composition in the following epochs. I might need to lower the shift during training to it focuses more on the details.