Edinburgh Stone Suite - Silk Camisole Candid & Krea 2 Turbo

1. ARCHITECTURAL REASONING & COMPOSITION GEOMETRY: Deconstructs an intimate candid portrait in an Edinburgh historic stone suite bedroom. While Krea 2 Turbo excels at photorealistic human form and macro skin rendering, standard checkpoints often struggle with complex textile textures like handwoven Scottish wool tartan and delicate lustrous silk camisole fabrics in the same frame. Chaining the Realistic Snapshot adapter at 0.8 weight injects authentic documentary photography characteristics: subtle film grain, natural pore micro-relief, unposed eye contact, and realistic fabric drape without artificial plastic smoothing. The Refusal Reduction V2 full-rank adapter at 1.0 weight preserves nuanced spatial instructions, preventing prompt clipping across layered bedroom props (mattress linen, tartan blanket fringe, sandstone wall window bounce). 2. CABLE ROUTING & SIGNAL TOPOLOGY: Sequential dual LoRA chaining topology over Krea 2 Turbo UNET (krea2_turbo_int8_convrot.safetensors) and Qwen3-VL 4B text encoder (qwen3vl_4b_fp8_scaled.safetensors). Model and CLIP channels pass sequentially through Refusal Reduction V2 and Realistic Snapshot V2 before feeding the conditioning nodes. A single deterministic Positive conditioning node is wired directly to KSampler, enforcing the WYSIWYG canvas invariant. The empty latent is sized at 896x1152 (3:4 vertical aspect ratio) and decodes cleanly through native qwen_image_vae.safetensors into RGB space. 3. HARDWARE BENCHMARKS & CONSUMER GPU FOOTPRINT: Executed on host NVIDIA GeForce RTX 4070 (12GB VRAM, CUDA 12.4). Peak VRAM usage reaches ~10.4 GB during the 12-step Euler simple generation, well within the 12GB threshold. Total inference latency completes in 70.1 seconds without host RAM swapping or precision degradation.