Img2img is short for image-to-image — a generation mode where the model is given both a text prompt AND a reference image, and produces an output that respects the structure of the reference while changing what the prompt asks. It's the workhorse mode of AI interior design.
The key parameter is "denoising strength" — how much of the original image to preserve. Low strength (0.2-0.4) produces small changes: a lamp swap, a paint color shift. High strength (0.7-0.9) produces a near-total restyle that only loosely follows the reference. Most interior design tools sit in the middle (0.5-0.7), which keeps the room's geometry recognizable while letting style, materials, and furniture change freely.
Img2img is how a user upload becomes a redesign. The alternative — pure text-to-image — has no anchor to reality, which is why prompts like "modern living room" produce generic stock-style outputs without your room's actual proportions.