ModelsLab/Astigmatism (Formerly 'Semantic Shift') - Astigmatism +0.6

Used my dataset from producing the flux version of this, which was a lot better curated than my previous attempts, to produce this SDXL version. It's substantially improved in terms of avoiding overtrain, but 0.5 remains special in terms of aspects of adherence, although it has issues when turned up with causing some types of artifacting, which this version does not have. Combined, they are beautiful.