From SketchUp to client board in one afternoon

The brief arrives on Tuesday and the client sees it on Wednesday. There is a SketchUp model of the massing, there is no visualiser, and there is no version of the old pipeline that ends before Friday. This is the afternoon that replaces it.

A note on what this is: a worked example rather than a named project. The practice is a composite of the ones we hear from most often, and every timing below is a published figure from the benchmarks page applied to it — not a stopwatch on somebody’s Tuesday. Your afternoon will be faster in the places where you are decisive and slower in the ones where the client is not.

Play AlphaRender walkthrough

13:00 — Export

The export is the only part of the afternoon that happens in SketchUp, and it takes about ten minutes if the model is tidy and about thirty if it is not. What matters:

• Frame the camera in SketchUp first. The viewpoint travels with the file, and choosing it in the tool you model in is faster than choosing it anywhere else.
• Export OBJ or FBX — up to 500 MB, which is far more than a massing model ever needs. No UVs, no textures, no material library.
• Delete what is not being shown. A neighbouring block modelled as a box is context; forty imported furniture models are forty megabytes of noise.
• Keep the ground plane. A building floating on nothing gives the run no site to sit the shadows on.

13:15 — The first run, in Creative mode

Upload, leave the presets alone, pick Creative, generate. One credit, about 68 seconds, 10–20 directions. The temptation at this point is to start tuning; do not. The first run exists to tell you what the model wants to be, and you cannot know that before you have seen twenty answers.

While it runs, start the second one on a different camera. Up to eight generations run in parallel, so the elevation view and the approach view are both waiting for you when you come back to the screen.

13:40 — Shortlist three

Three passes, about three minutes: squint at the grid for value and silhouette, open the survivors and check the ground, glass and main material, then keep the ones that answer the sentence the client actually said. The full method is here. You are looking for three arguments, not three favourites — the literal reading of the brief, the one you would build, and the outlier.

14:00 — Fix instead of re-rolling

This is where the afternoon is won or lost. Every direction you keep will have one thing wrong with it, and the instinct is to generate again and hope. Three of the four common faults are edits, not runs:

1. The hour is wrong

   Relight the frame you already have rather than rolling for a better sky. Presets carry the whole day, and the shadows move with them — the lighting lesson is the interior version of the same tool.

2. The site is wrong

   Move the project instead of the camera. The environment tool swaps the surroundings under a finished render, which is also how you show a client the same design on two plots.

3. It is empty

   A public building with nobody in it reads as a model. Add people at the density the brief implies, and add them at the right distance from the camera.

4. The design is wrong

   This one is a real re-run, and it is good news — you found it at two in the afternoon on a one-credit render instead of on Friday after a farm bill.

15:00 — The revision round

Send the three to whoever in the practice has to sign them off, and treat what comes back as one round. A ten-option revision round measures at 41 minutes and $6.60 — against one to two days on a render farm, and one to two weeks through an outsourced studio. That gap is the actual product: not the first image, the second one.

41 min

a ten-option revision round

$6.60

what that round costs

≈ 60

concepts explored per project

16:00 — Upscale, and only then

Now the frames are settled, spend credits on them: 4K is the right size for a board, a slide or an A3 panel, and 8K is for the printed spread. Upscale the three you are showing and nothing else — what actually changes at each size is worth two minutes of reading before you spend twenty credits.

17:00 — The board

• Same camera across all three. The board is about the building, not about the photography.
• Label each frame with the decision, not the preset: “Evening, brick” tells a client what they are choosing.
• If one image has to carry the meeting, generate ten seconds of video from it and lead with that.
• Keep the twenty you rejected. When the client asks whether you considered a lighter facade, the answer is a folder, not a promise.

“This is the exact solution I have been looking for in my architectural practice: a fast, precise, and frictionless workflow that delivers production-ready quality.”

Wang Khanh, Tech Lead, vacs.co.jp

What the afternoon cost

Four runs at one credit each, one revision round, three 4K upscales. On the Basic plan at $99 a month that is a rounding error against a single day of a visualiser’s time, and the plans page has the arithmetic for teams. The part that is not on the pricing page is the one that matters: the client saw three directions on Wednesday instead of one on Friday.