I was just stating the improvement on SD we've seen since DELL-E 2 and Midjourney came out wasn't just about "quality of image", but also about "have something specific and moderately complex in mind". Thus I mentioned textual inversion vs LoRA/ControlNet.