3D vs photorealistic landscape rendering

"Rendering" arrived in yard design from two directions at once. Architecture used it for the lit camera view produced at the end of a 3D modelling job. Print and drafting used it for colouring up a line drawing so a client could read it. Then photo-editing tools started producing photorealistic images with no drawing underneath at all, and kept the word.

So the term now covers a model render, a photo edit and a coloured plan. A homeowner asking a designer for "a rendering" usually means the photo edit. A designer hearing it usually costs the model render. A contractor pricing from it needs the plan, and cannot work from either of the others.

The gap between those three is thousands of dollars and several weeks. Our landscape rendering overview covers the pricing bands; this page is about telling the deliverables apart before you buy one.

This is the one architects mean. Somebody measures the site, builds it as geometry in modelling software, places the structures and planting, then points a virtual camera at it and renders a frame.

The model is the actual work; the image is a by-product. Because the geometry exists, you can spin the camera to any angle, cut a section through a retaining wall, take dimensions off it and produce an elevation. Sun and shade studies work because the software knows where north is and how tall the wall is.

The tooling reflects that. SketchUp Pro is listed at $33.25 per month billed annually, with SketchUp Studio at $68.25 per month, on SketchUp's plans and pricing page read on August 21, 2026. Lumion, which renders a model somebody else built, lists Lumion View at $229 a year and Lumion Pro at $1,149 a year on its buy page, also read August 21, 2026.

This is the kind most homeowners are picturing. You photograph the yard, the tool redraws the planting and paving inside that photograph, and your house, fence line, driveway and neighbour's roof stay where the camera saw them.

There is no geometry underneath. Nothing was measured. The output is one frame from one viewpoint, and there is no model to rotate, section or dimension. What you get in exchange is speed and price: AI Landscape Design returns a redesign of your own photo in about 10 seconds across 18 styles, with the first 3 free at full resolution.

The realism is genuine — planting texture, paving joints, shadow direction — which is exactly why it gets confused with a model render. The two look similar and guarantee completely different things. One came from measurements. This one came from pixels.

The third kind is the one that gets built from. It is an overhead drawing of the site with colour and texture laid over it so a non-drafter can read it: beds shaded, paving hatched, trees drawn as canopies, everything to scale.

The NC State Extension Gardener Handbook describes the base plan as "a bird's eye view of the site drawn to scale" and recommends one inch to ten feet for an average home yard, with a north arrow marked on the drawing — see its chapter on landscape design. Because it is to scale, quantities come off it: square feet of paving, linear feet of edging, plant counts.

It is not a picture of the finished space, and nobody has ever been persuaded by one. It is a working document that a contractor prices and a crew installs from. Vectorworks Landmark, which produces exactly this alongside grading and site models, lists at $1,530 a year on its buy page, read August 21, 2026.

The limits are not preferences, they are consequences of how each thing is made.

A photo edit cannot be rotated, sectioned or measured, because there is nothing behind the image to rotate. It cannot read a slope, because a single photograph carries no elevation data — sloped yards routinely come back rendered flat. It cannot price anything, because no quantities exist in it.

A 3D model render can do all of that and still cannot tell you what a paver costs in your county, or whether your city permits that fence height. It also costs weeks: somebody has to measure and model the site first.

A rendered 2D plan carries dimensions and quantities and is close to useless for persuasion. Show a spouse an overhead diagram and you will not settle the argument you were trying to settle.

Work backwards from what happens next.

If the next event is a decision — which style, how deep the beds, whether the pergola belongs — you need a photo edit and nothing else. It is the only one of the three that shows your actual house, and it costs nothing to try. Run your own photo through AI Landscape Design before spending anything.

If the next event is a quote or a build, you need a plan. Dimensions, quantities and a plant schedule are what a contractor prices; a picture is not.

If the next event is a planning committee, a client presentation, or a build involving retaining walls, steps and grade changes, you need the 3D model. That is the only version that survives being questioned about levels.

Most residential jobs need the first two and never the third.

AI Landscape Design is a type two tool and only a type two tool. The list of things it does not do is worth stating plainly, because every one of them is something buyers assume a "rendering" includes.

It produces a concept image, not a construction document. It does no grading and no drainage analysis. It takes no measurements and outputs none. It knows nothing about setbacks, easements, HOA rules or permits. It cannot be rotated to a second angle, because there is no model — a new angle means a new photograph.

It also draws planting at mature size, so what you see is roughly year five, not planting day. Download it for the job it does, which is deciding what you want. For the other two jobs, hire the other two deliverables.

Buy them cheapest first, because each one narrows what the next has to cover.

Start with photo edits. At about 10 seconds and no cost for the first 3, you can rule out four directions in an evening. Ruling things out is most of the value: a style discarded here is a style nobody drafts.

Then commission the plan, with the render in hand. A designer given "this look, on this yard" bills less than one asked to start from a conversation, because the exploratory work already happened.

Only then consider a 3D model, and only if levels, walls or a committee force it. At $1,149 to $1,530 a year for the software alone, plus the modelling time, it is the expensive option and most yards never justify it. Start at the free end.