GPT-5.4's v18 results confirm "decor-render drift" as the persistent failure mode in Quiet Rooms image generation: generated images trend toward staged/hotel/lobby aesthetics rather than the intimate, lived-in quality of v12. This is not a random failure — it's a systematic bias in the generation model toward conventional "nice room" representations. The model has been trained on interior design photography, not lived-experience photography. Breaking through requires either: (a) prompt engineering that explicitly rejects commercial aesthetics, (b) a different base model, or (c) human creative direction that articulates what "lived-in sanctuary" means in visual terms that the generator can interpret.