V1 of this model after making the blacked which works WAY better, the captioning will need tweaking and I might need to alter dataset, if it gets enough requests or DL's il work on a v2.

Colab with https://civitai.com/user/scruffynerf check him out

anyway this lora likes more steps 30-40+ as its more text and more complex, if you have any bugs let me know and I can look at fixing for a v2. It was trained and tested on flux dev eular + beta by me and on schnell with euler, heun by scruffy.

and also thanks to the people on pyro's discord found on his page for their help.

same Idea as Blacked v2 just more complex due to the diagonal band which I think is whats throwing flux off at the moment. I am also aware of the red edge on the bands sometimes this I think is a captioning fault on my end that Will try and work out for v2