JedPoint model completely repeats the base model (flux1-dev-bnb-nf4-v2, flux1-dev-fp8,), but yes when using the code word “Jedtoon” - small differences appear, but the word itself - can come out in the picture. I think the model is useless for now, the changes are small. The author thanks, of course - but somehow you need to train it to be different from the standard and not use the stop word “Jedtoon” .