Trained from a modified dataset based on my SD1.5 LoRA and I'm quite happy with the result no matter if Flux just did its magic or if the added input images did contribute as much as I'd hoped.
First release