Thanks guys for the feedback. I’ve changed the preset generation model from GPT-5 Mini to GLM-5. The quality has improved a little, but the model is slightly slower