## The short answer
Midjourney makes images from prompts and reference pictures. It is very good at mood, atmosphere and ideas, and many architects use it for concept boards. It does not know your 3D model. Each image is a fresh interpretation, so it will not reliably give you the same building from three cameras. A model-aware tool such as OctopusWave starts from a view of your model and keeps the building and the camera, changing the surfaces, the light and the setting. Use the first when you are exploring, the second when you are presenting your design.
## What Midjourney is great at
Be fair to it. For the first hours of a project it is hard to beat:
- A fast way to find a mood: coastal brutalism, a warm timber pavilion, a misty valley.
- Strong images from a short sentence, which makes it good for brainstorming with a client.
- A style reference option that helps a set of images feel related.
It is a general image tool, and that is its strength.
## Where it stops being the right tool
- **It does not take your model.** You can show it an image, but it treats that image as an inspiration, not as geometry to preserve. The number of floors, the position of the cantilever and the lens can change from image to image.
- **Each image is independent.** A street view, a courtyard view and an aerial of the same project will not agree about the building.
- **No link into your modelling software.** As far as we know at the time of writing, there is no official Midjourney plug-in for SketchUp, Rhino or Revit, so every image starts with a screenshot and a manual upload.
- **No design documents.** You get pictures, not a deck, a film from the renders or a first draft of a design statement.
None of this is a flaw. It is what the tool is for.
## What a model-aware tool does differently
OctopusWave starts from your view. The plug-ins for [SketchUp](/plugins/sketchup), [Rhino](/plugins/rhino) and [Revit](/plugins/revit) send the current view, or every scene or named view at once. The render keeps the volumes, the proportions and the camera. Then:
- The same project renders in one style across all views, so plans, sections and street views belong together.
- Day, dusk and night versions come from the same camera.
- A reference image gives the render its materials and photography without redrawing your building.
- A personal style trained on your own images, [My Signature](/news/train-your-own-ai-render-style-my-signature), makes the studio look consistent.
- Renders can go on into a short film, a 360 panorama or a PDF deck.
## A simple test to know which job you have
Ask one question: **does the building have to match the model?**
- If the answer is no, because you are looking for ideas, use Midjourney or any general image tool.
- If the answer is yes, because this goes to a client, a planning application or a competition board, start from the model.
## How to use both
The workflow many studios settle on:
1. Explore the mood with a general image tool.
2. Save the two or three images that say what the project should feel like.
3. Model the massing.
4. Send the views to a model-aware tool and use your saved images as the **reference**, so the render takes the mood while the building stays yours.
That gives you the creative range of the first tool and the accuracy of the second.
## What to check in any AI render
Whichever tool you use, check each image for the same things: the number of floors, the position of openings, the repeated elements such as mullions and balusters, and any text. AI images are convincing, which is exactly why a slip in the details is easy to miss.
## Try it
New OctopusWave accounts get five free renders from the plug-ins. Send one real view from your model and compare it with what a general image tool gives you for the same project. After the free renders, Model Link is part of the [Studio plan](/pricing), or you can pay per render with credits and no subscription.