## The Composition Problem
Standard AI image generation models are trained to produce aesthetically pleasing outputs — not structurally accurate ones. Feed them an architectural sketch and they will generate something beautiful. It will rarely match what you drew.
Window positions shift. Proportions change. The structural logic of the building — the thing that makes it your design — gets overwritten by the model's aesthetic preferences.
For architectural practice, this is not a minor inconvenience. It is a fundamental incompatibility.
## ControlNet: Structural Conditioning
ControlNet is a neural network architecture that runs alongside the primary image generation model, conditioning the output on structural information extracted from your input. OctopusWave implements two ControlNet variants:
**FLUX Depth** extracts a depth map from your input — a pixel-by-pixel representation of spatial distance. The generation model is conditioned to respect this depth structure, preserving the three-dimensional organisation of your design: what is in front, what is behind, how volumes relate to each other in space.
**FLUX Canny** extracts edge information — the structural lines that define your composition. Walls, openings, rooflines, columns. The model generates within these constraints, producing photorealistic output that follows your drawn geometry.
## The Strength Slider
OctopusWave's Precision Render node includes a structural fidelity strength slider — a parameter controlling how strictly the output adheres to the conditioning input.
At maximum strength, the output is a direct photorealistic interpretation of your sketch: every edge preserved, every proportion maintained. At lower strength values, the model has more creative latitude — useful when you want the AI to suggest material or atmospheric variations while preserving the broad compositional logic.
## Two Input Ports
The Precision Render node accepts two inputs:
**Main input:** Your sketch or reference image — the design to be rendered
**Style reference:** An optional second image providing aesthetic direction — a material reference, a precedent project, an atmospheric photograph
The model synthesises both: your structure, their aesthetic.
## When to Use Precision Render
Precision render is the right tool when the composition matters — when you are presenting a specific design to a specific client and cannot afford for the AI to reinterpret it. For pure ideation and exploration, the standard render engine offers more creative latitude and faster generation.
For client-facing deliverables, planning submissions, and design development work where accuracy is non-negotiable, Precision Render is the appropriate choice.
## Getting Started
Precision Render with ControlNet conditioning is available on Pro and Studio plans. Add a Render node to your canvas, select a ControlNet model from the model picker, and adjust the strength slider to your required fidelity level.