Updated on: 30 October 2026
Realistic AI renders come from four things, in this order: a clean input view, a short prompt that names materials, light and surroundings, the right light setting, and fixing one area at a time instead of starting over. The ten render tips below work with any image-to-render AI; where a step needs a button, it names the control in ArchiVinci as of October 2026.
1. Start From a Clean View
The AI reads everything in the image you upload, so whatever is not part of the design can end up in the render: guide lines, axes, grids, section planes, selection highlights, toolbars or a cursor. Hide them before you capture the view. In SketchUp, for example, selected guides are hidden with Edit > Hide or removed with Edit > Delete Guides, the axes are hidden from their context menu, and File > Export > 2D Graphic saves the view as a JPEG or PNG at the pixel width you set, with anti-aliasing.
ArchiVinci takes one image per run: a JPG, PNG or WebP file up to 5 MB, such as a screenshot or an export of a view from any 3D, CAD or BIM tool, a sketch or a photo. It reads images, not model files (.skp, .rvt, .dwg and the like), and it has no scene, material or light setup, so the view and a few words carry everything. A plain, untextured view is enough: the prompt then decides the materials.
Left: the uploaded image. Right: the result in the Exterior module of the prompt "Modern house with white plaster walls, timber cladding on the upper floor, large glass openings, a lawn and a few trees." with the Daylight button, on the default engine.
2. Fix the Camera Before You Render
An image-to-render AI keeps the viewpoint it is given: it cannot walk around the building or find a better angle. Set the camera in your modelling tool first. Keep vertical lines vertical: SketchUp's Two-Point Perspective gives a view with two vanishing points, so walls do not lean. Use a natural standing eye height for street and room views, and a field of view that is not extreme: SketchUp sets it under Camera > Field of View, in degrees or millimetres, and a wider field of view helps inside rooms.
If the result needs a different angle, export a new view and render again. Asking the prompt for "a view from the garden" makes the model invent geometry it has never seen.
3. Pick the Module That Matches Your Image
There is no input-type switch to set: the module tells the AI what kind of picture it is looking at.
Exterior: a building from outside, as a model view, a massing view, a sketch or a photo. Its optional Add object tags add a Tree, People or a Car.
Interior: a room. Its Add object tags are Seating, Plants, People, Pet, Lamps and Accessories.
Floor plan: a 2D plan, rendered in one of four styles (3D view, Marker, Solid color or Drawing). Interior and Exterior are not meant for plans.
Masterplan: a site plan seen from above.
Imagine: a picture from a typed description alone; write it in English.
For drawings, the sketch to render page shows the route from line work to a finished image, and the AI render generator page walks through Exterior and Interior with examples.
No Model Yet? Start From the Floor Plan
If the project exists only as a plan, Floor Plan Maker gives you a 3D view to render. In a desktop browser you draw the plan from scratch, or upload a plan image, a PDF or a hand sketch for AI to rebuild as an editable plan (1 credit on a free account, refunded if the conversion fails). Add walls, doors, windows and furniture from 153 pieces (wall lengths and furniture dimensions appear as you draw, like in a CAD program), check the rooms in the 3D view or in a first-person walkthrough, then render the current view as an interior or exterior scene with ArchiVinci's engines (normal render credits), with a lighting choice and an optional prompt. Drawing, the 3D view, the walkthrough and the exports are free.
No model yet? Render from your floor plan Draw the plan in Floor Plan Maker or upload the one you have, furnish it, check it in 3D or in a walkthrough, then render the current view as an interior or exterior scene with ArchiVinci's engines. Drawing, 3D and exports are free in a desktop browser; reading an uploaded plan with AI costs 1 credit on a free account. |
4. Describe Materials, Light and Context in One or Two Sentences
In Exterior and Interior the prompt is optional: with no prompt, the AI renders what the image shows. When you write one, spend the words on what the image cannot say: the materials, the light and the surroundings. Keep it to a sentence or two. The app warns above 200 characters, and a short prompt with one intent is easier to check against the result.
Words that make a render read as real are concrete. Name the material and its finish (white plaster, board-formed concrete, oiled oak, brushed steel), the light (soft overcast daylight, low warm sun from the left) and the context (a lawn and a few trees, a wet street, neighbouring buildings). Leave out words that describe image quality rather than the building, such as "8K", "ultra-detailed" or "masterpiece", and do not mix two styles in one sentence. Examples:
Exterior, from a sketch: "Add clouds, grassy ground, soft shadows."
Exterior, from a plain view: "Modern house with white plaster walls, timber cladding on the upper floor, large glass openings, a lawn and a few trees."
Interior: "Warm modern living room with an oak floor, a beige linen sofa, sheer curtains and soft daylight."
For a fuller formula with 15 examples by building type, see the guide to AI architecture prompts.
Put these render tips to work on your own image Upload a sketch, a 3D model view or a photo, describe materials and light in a sentence, and ArchiVinci renders it as a photorealistic image in your browser. A new account starts with 3 free credits. |
5. Set the Light With a Button, Then Fine-Tune in Words
Exterior and Interior have a Lighting (optional) row with six buttons. In Interior they are Daylight, Warm light, Twilight, Golden hour, Ambient and Studio light; Exterior's six include Daylight, Golden hour and Twilight. A button changes the light without any change to the prompt, so it is the quickest way to compare moods on the same view. Add light words to the prompt only for what a button cannot say, such as "sun from the left" or "pendant lamps on".
The same upload and the same prompt as the house above, in the Exterior module on the default engine. Left: the Golden hour button. Right: the Twilight button.
Twilight gives dusk with the lights on; no button gives a full night scene. For a night version, upload the result to the Edit module (or pick it from the render history) and type the request: the page on how to relight a render shows the steps.
6. Keep the Engine on the Default for Fidelity
The Render engine menu sets which AI model makes the image. The default, ArchiVinci 4.0, keeps the perspective, geometry and colours of your upload. It uses 1 credit per image on a free account, and every plan includes unlimited renders on it. As of October 2026 the same menu also offers partner models (Nano Banana 2, GPT Image 2, Seedream 5.0 Lite and Qwen Image 2) at 1 coin an image; free credits do not pay for them, and they are there to see how a different model reads the same view. When fidelity to the design matters, start on the default.
7. Generate Several Versions, Then Choose
There is no similarity slider, strength setting or seed: every run generates a new image that follows your upload, so two runs of the same view and prompt never come out identical. Use that. If a result is close but not right, run it again before you rewrite the prompt. With a plan you can generate up to 4 images per run (one per run on a free account) and pick the best. Every result stays in the render history for 7 days, so download the ones you keep.
8. Change One Thing at a Time in the Edit Module
When one chair, one material or one window is wrong, do not render the whole image again: a new run changes everything else too. The Edit module, which ArchiVinci calls ArchiGPT, changes an image by a typed request, and there the prompt is required. Without a marked area it works on the whole picture. To change only part of it, paint over that part with Select (a brush) or outline it with Select area (a lasso), press Done and describe the change. Under a result, Edit render does the same on that result: mark the area and replace the old prompt with the new request.
To remove an object, mark it and ask for its removal ("remove the chair"); the Remove button under the image only erases part of your selection. One product picture can go under Upload object (optional), and the model works it into the image. Apply texture puts one of 103 library materials in 15 categories, or a photo of your own sample, on a marked wall or floor. Only the marked area changes and the rest of the image stays as it is. The AI image editor page shows the module with examples.
9. Change the Style Without Losing the Design
There is no style picker and no style-strength slider: the style is words in the prompt. On a plain, untextured view, name the style with its key materials in Exterior or Interior ("Scandinavian living room with light oak, white walls and wool textiles"). A render or photo that already has its finishes is restyled in Edit: write the change out part by part ("walls in white lime plaster, floor in light oak, keep the furniture") rather than one word like "make it modern". Every run draws a new variation, so run it again for another option or mark one part to refine it. A picture under Upload object is sent to the model as an object to add, not as a style reference.
10. Finish: Upscale and Animate
Under a result, Upscale enlarges the image to up to 4096 px on the long side, as a JPEG. It needs a plan and works only on images generated in ArchiVinci, so a render from another program goes through one Edit pass first. Fix wrong details in Edit before you upscale; AI upscaling explains what an upscaler adds and what it cannot repair.
With a plan, Animate under a result and the Video module turn a render into a clip. The camera move is a control, not a prompt: None, Pan, Zoom in, Tilt or Zoom out, with a direction for Pan and Tilt. Every video is 10 seconds long; it starts from one frame, or from a start and an end frame (Start & End frames); the quality is Normal, High or Best, paid with the coins every plan includes. The prompt is optional: if you write one, describe motion that fits the chosen move, such as "leaves moving in a light breeze, people walking past", not a second camera move. The page on how to turn a render to video has examples.
Common Mistakes That Make Renders Look Fake
Guides, grids or a cursor left in the view. The AI can turn them into lines on walls or in the sky: hide them before you export.
Leaning verticals. A tilted camera makes walls converge; use a two-point perspective for exteriors.
Long prompts that mix styles. Two styles or ten materials in one sentence give a muddled result; keep one intent per run.
Asking for a new viewpoint. The model has to invent what it never saw; export the view you need instead.
Expecting a night scene from a button. Twilight is dusk with the lights on; type a night request in Edit.
Rendering the whole image again to fix one chair. Mark the chair in Edit and change only that.
Floating or merged details. Check columns, mullions, railings and where furniture meets the floor before you send the image, and fix a bad spot with a marked edit. For the craft behind convincing light and materials, see photorealistic rendering.
Try it in ArchiVinci
Upload a sketch, a photo or a 3D model view and describe the result in a sentence: ArchiVinci generates a new image that follows your upload. A new account starts with 3 free credits.
From a photo or an image: Render an image with AI
For a render you already have: Enhance renders with AI
Render Tips FAQ
Why does my render look fake?
Usually for one of three reasons: the input carried things that are not part of the design (guides, grids, a cursor), the camera was tilted or extreme, or the prompt described image quality instead of materials and light. Clean the view, straighten the verticals and name the materials, the light and the surroundings in a sentence or two.
How do I keep the render true to my 3D model?
Upload a clean export of the exact view, keep the Render engine on the default ArchiVinci 4.0, which keeps the perspective, geometry and colours, and use the prompt for materials and context rather than for changes to the building. To refine a detail, mark it in Edit and describe the change.
Why does the AI add objects that are not in my model?
Every run generates a new image, and an image model tends to fill empty space with what such scenes usually contain. Say what should stay empty ("empty terrace, no furniture"), leave the Add object tags off unless you want those things, and remove a leftover by marking it in Edit and asking for its removal.
How long should a render prompt be?
One or two sentences. In Exterior and Interior the prompt is optional and the app warns above 200 characters; in Edit it is required. Spend the words on materials, light and context.
Can I get a night render?
Not from a lighting button: Twilight gives dusk with the lights on. For a night view, upload the result to the Edit module (or pick it from the render history) and type the request, for example "night scene, interior lights on, dark blue sky".
Can I use free renders commercially?
No. A new account starts with 3 free credits to try the app, but using the images for client work, marketing or sale needs an active plan, which includes commercial use rights. Plans are on the pricing page.