Midjourney generates a beautiful concept image in 30 seconds. The proportions look right. The atmosphere is perfect. The materials feel real. Your client approves it.
Then the image goes to the drafting team — and it falls apart.
The concept looks one way in the image and a different way when drawn as a technical plan. The magic is lost in translation. The client notices. The interior designer is caught between a concept they cannot defend and a technical team they cannot brief. Something went wrong.
Nothing went wrong. This is what was always going to happen.
Why Midjourney Cannot Produce Buildable Images
Midjourney operates on probabilistic pixel distributions. It generates what looks right visually. It has no understanding of structural engineering, building regulations, or construction feasibility. Its outputs are 2D pixel representations of imagined 3D spaces — visually compelling and geometrically unconstrained.
A Midjourney image might show a room with a ceiling that appears to be 8 metres high. The model has no idea whether that ceiling is structurally possible, fire-safety compliant, or requires engineering input that the image simply cannot show. This is not a flaw. It is the nature of the tool.
When that image is handed to a drafting team with no accompanying dimensions, one of two things happens:
The drafting team tries to reverse-engineer proportions from the image. They estimate ceiling heights from the apparent scale of furniture. They infer structural grid spacing from the spatial impression. They get it wrong — because the proportions in the image were never measured, never verified, and never intended to be taken as dimensions. The resulting drawings look subtly wrong compared to the approved concept.
The drafting team ignores the image. They produce technically correct drawings that are creatively flat — accurate to every building regulation, completely devoid of the spatial character the client approved.
Neither outcome serves the designer, the client, or the project.
The Client Expectation Problem
Clients approve Midjourney images because they look amazing. They do not understand — and are not told — that what they are approving is a creative direction, not a technical specification.
This is a professional liability that AI has made worse. Concept imagery is now more persuasive and easier to approve than ever before. When the drafting team produces technically correct but visually different drawings, the client feels that something has been lost. They approved Option A. The drawings don't look like Option A.
The solution is not to stop using Midjourney. The solution is to reframe what the image is and what it is not — from the outset.
The Epistemic Distinction Every Designer Must Make
Before using Midjourney professionally, the interior designer must understand one distinction clearly: Level of Detail (LOD).
Midjourney operates at LOD 0–100: conceptual spatial impressions. CAD and BIM drafting operates at LOD 200+. These are not the same document. They are not different versions of the same document. They are different categories of document.
When a client approves a Midjourney image, they are approving a creative direction — not a technical specification. The interior designer's job is to make that distinction explicit before approval is sought, not after it is given.
Every Midjourney output must be labelled: "AI Draft — concept imagery for client presentation only, not a buildable specification." This is not a legal formality. It is a professional communication standard that protects everyone in the process.
What AI-Generated Imagery Cannot Tell You
No matter how photorealistic a Midjourney image, a 2D pixel representation cannot reveal:
- Exact measurements (all dimensions are estimates)
- Structural feasibility
- Building regulation compliance
- MEP (mechanical, electrical, plumbing) locations
- Thermal performance
- Sound insulation values
- Site-specific constraints
For projects in England and Wales, the building regulations most commonly relevant to interior design spatial parameters are Approved Document B (fire safety — minimum ceiling heights, fire door sizes), Approved Document M (access — door opening widths for wheelchair access), and Approved Document K (protection from falling — mezzanine guardrail heights, minimum 1.1m). None of these can be inferred from a 2D concept image.
The Three Numbers That Define What AI Can and Cannot Do
The interior designer's epistemic framework for AI-generated imagery rests on three numbers:
LOD 0–100: Midjourney. Concept. No dimensions, no structural data, no compliance. Use for client communication and creative exploration only.
LOD 200–300: ChatGPT with Advanced Data Analysis. Geometric estimation from a 2D image using human scale references and UK architectural conventions. Produces a starting point for professional review — not a verified specification.
LOD 300+: CAD and BIM. Working drawings, structural engineering sign-off, building regulation compliance. The document that instructs a contractor what to build.
The gap between LOD 0–100 and LOD 300+ is where interior design projects most commonly fail to deliver on the creative vision. The concept-to-translation problem is real, and it is not solved by better prompts or more photorealistic rendering. It is solved by a structured workflow that manages the transition between levels of detail — with the designer authorising every dimension at every step.
Why the Interior Designer Is Always the Author
AI accelerates; the designer authorises.
This is the principle at the heart of every WIZGUIDES guide, and it applies here with particular force. Midjourney generates imagery. ChatGPT estimates proportions. The interior designer reviews every value, checks it against building regulations and structural engineering input, and authorises the final spatial brief before the drafting team receives it.
AI does not replace professional judgement. It extends the time the professional has to apply it. In a traditional workflow, the interior designer spends three to five days on manual spatial proportioning and repeated sketch rounds before the drafting team has enough to work with. With the structured constraint sheet workflow, that same starting point is reached in 45 minutes — leaving the designer with more time to review, correct, and refine.
The designer who understands this distinction will use Midjourney as the most powerful concept communication tool in their workflow. The designer who does not will spend their career apologizing for drawings that don't look like the approved concept.
Related Resources
The gap between AI concept imagery and buildable specifications is the specific problem From Generative Visuals to Geometric Parameters was written to solve. The five-step constraint sheet workflow uses ChatGPT with Advanced Data Analysis to extract estimated spatial proportions from a Midjourney concept image into a structured CSV format — reviewed and authorised by the designer before the drafting team receives it. It includes confidence levels (HIGH / MEDIUM / LOW) for every estimated value, a full UK architectural conventions reference table, and the Technical Translation Brief template that closes the gap between approved concept and working drawings.
If you are already working in the architectural space and using AI for concept exploration, you are likely also dealing with the consistency problem in image generation — generating the same space with different materials and getting completely different rooms. The Controlled Iteration Framework for AI Spatial Prototyping covers the four-step workflow for solving this using Midjourney's `--sref`, `--cref`, `--seed`, and `::weighting` parameters. It is the companion guide to this one.
Stop Guessing, Start Systemising
The concept-to-translation problem is not a Midjourney problem. It is a workflow design problem. Every interior designer who uses AI-generated imagery professionally needs a system for managing the transition from LOD 0–100 to LOD 300+.
That system is documented in full in the From Generative Visuals to Geometric Parameters guide. It includes the complete five-step workflow, the CSV constraint sheet template, the UK architectural conventions reference table, and the Technical Translation Brief structure. Forty-five minutes of structured workflow instead of three to five days of manual proportioning.
Grab it here