Camera composition for archviz stills: what actually makes an image work

When a client only pays for four hero images, the composition of each one has to do more than just show the building. It has to convince someone who has never been to the site that they know what it feels like to stand there. That is a much taller order than most junior artists realise, and it is the reason a technically perfect render can still feel flat. I want to talk about how I choose where to put the camera, because it is the least discussed part of the pipeline and probably the most important.

Start from the plan and section, not from the 3d view. Before I move the camera in the software I sit with the drawings for 15 minutes and mark five potential vantage points, three exterior and two interior. For each one I write down what the image is supposed to communicate. Exterior one might be the street presence and material palette. Exterior two might be the outdoor living relationship. Interior one is the main living volume and the light from the primary window. This is planning work and it saves me from the trap of just orbiting until something looks nice.

Camera height is the setting people mess up most consistently. Eye level for a standing adult is 1.55 to 1.7 meters. Not 1.2, which is where SketchUp likes to put the camera by default and which reads as a child's view. Not 2.5, which reads as slightly elevated drone. For exteriors of houses I use 1.65 nine times out of ten. For interiors I sometimes drop to 1.4 because a seated viewpoint reads as more inhabited, less like a real estate listing. Test both and see which one makes the space feel right.

Lens choice is the next lever and there are really only three lenses I use. A 24mm equivalent for tight exterior sites where I need to fit the building and some foreground context. A 35mm for interiors and for exteriors where I have room to step back. A 50mm for interior detail shots and for exterior close ups on a facade material. I basically never go wider than 24mm because it distorts too much and starts to feel like real estate video. I basically never go longer than 85mm because it flattens the space so much it feels like an elevation drawing.

Verticals should almost always be corrected. This is the one photography rule that translates directly. In V-Ray, Corona, or Blender's camera settings, look for the option to make the camera vertical shift or tilt shift. Turn it on and the building lines that would otherwise converge upward stay parallel. This is what architectural photography does with tilt shift lenses. Skipping this step is what makes amateur renders read as amateur before anything else in the image gives it away.

Foreground matters more than the building. This sounds heretical but I promise it's true. A render with the building perfectly framed and nothing in the foreground reads as a technical drawing. The same render with a bit of planter in the corner, a person walking through, or even just a shadow from an offscreen tree, becomes an inhabited place. I always populate the foreground even if it means moving the camera three meters to include something. The building is the subject but the foreground is what tells the viewer they are looking at a real world scene.

The last thing, and this is the one that separates people who ship work from people who don't, is knowing when a composition isn't working and moving on. If I've spent 40 minutes fiddling with a camera and it still feels wrong, I abandon that vantage point and go to the next one from my plan. Composition is not something you can force. Some sites just don't have the shot you thought they would, and pushing through with a bad composition wastes another two days of lighting and materials on an image that will never land. Trust the plan you made at the start, and if none of the five vantage points work, sit with the drawings again.