Nice job. FWIW, I honestly think this would have been more useful back in the days of SD 1.5 and SDXL, when prompt witchery was far more important. Nowadays, SOTA models like NB Pro and gpt-image-1 are really good at adhering to even moderately complicated prompts expressed in plain natural English.
Nice job. FWIW, I honestly think this would have been more useful back in the days of SD 1.5 and SDXL, when prompt witchery was far more important. Nowadays, SOTA models like NB Pro and gpt-image-1 are really good at adhering to even moderately complicated prompts expressed in plain natural English.