One-image model comparisons are misleading. I use a repeatable test: lock the prompt and reference set, vary one setting, then score prompt adherence, composition, readable text, reference consistency, editability, and credit/time cost. For Midjourney V8.2, compare V8.1 baseline → V8.2 baseline → V8.2 + personalization → V8.2 + SREF; Omni Reference remains a V7 workflow. Full checklist: https://www.pixmind.io/posts/midjourney-v8-2-release #ai #design