Generate Renders

Every render is the product of: the uploaded view, a model, a resolution, a time of day, a location, a camera angle, a style, and an optional prompt. This page covers each of them, ending with Continue at the bottom of the panel.

📺 Prefer to watch? See the full walkthrough on YouTube (in Spanish):

Where to start

There are three ways to reach the panel, depending on where you begin:

• New project: you upload your floor plan, sketch, or drawing, and that view is ready to render.
• Existing project (no 3D): you open the project and click one of its views to start rendering it.
• From the 3D Scene: move around the scene with drone or orbit mode until you find the shot you want, then click Render to capture that perspective.

All three land you in the same place: the render configuration. Don't obsess over the perfect framing yet; you can save several perspectives and render them one by one.

Render configuration

You don't have to touch everything. These are the decisions that actually change the result:

• Model. Exploring or delivering? Standard to try ideas fast and cheap; Pro when you want client-ready photorealism. Simple rule: the first render of a project, in Pro, because it gives the AI the most consistent base.
• Resolution. Sets the output size of the render: 1K, 2K, or 4K. Higher resolutions carry more detail and cost more credits. Standard supports all three; Pro starts at 2K (no 1K). See How credits work for the exact per-render cost of each model and resolution.
• Location. If you inserted your 3D model and set the location, it's there by default. Otherwise, select it.
• Time of day. Sets the mood of the scene. Noon is the most neutral; sunset and dusk, the most cinematic. Golden hour almost always looks great on exteriors.
• Camera angle. Keep your shot or try another? Same as drawing preserves exactly the perspective you captured (the most common). Custom if you want the AI to propose a different angle.
• Style. Sets how it looks. Real photo for photorealism, or upload your own reference: that Pinterest image with the mood you're after.

Add more details

This is the field where you tell the AI what the model can't say on its own. How to make the most of it:

• Short, always. One or two lines are enough. I know it's tempting to write the perfect paragraph (or ask ChatGPT for it), but here the opposite happens: long prompts, especially ones built with another AI, tend to make the render worse. They fill up with instructions that contradict each other and pull the image away from your model.
• Describe the abstract, not the obvious. The geometry, the measurements, and the openings are already in your model, so don't describe them. What adds value is the intangible: the atmosphere, the feel, "make it feel like a quiet neighborhood", "exposed concrete facade", "warm end-of-day light". You bring what the AI can't guess by looking at the model.
• If you can't say it in words, show it. When a material or mood is hard to describe, upload an image under References and point the prompt at it: "use the uploaded image as the facade material". A good reference is worth more than three lines of adjectives.

References

The References window has four tabs: All (everything together), Materials (a photo or sample of a material, which you can pair with Select area), Styles (images that capture the look you're after), and Objects (furniture, plants, and props to insert). Each has an Upload button to add your own images, plus a library of examples.

Continue and the questionnaire

You don't need to write a perfect prompt: the AI completes it by asking you questions. When you click Continue, it reads your reference and your model, and asks you two or three short questions (the surroundings, the light, the materials), each with buttons to pick with one click.

• Don't care about one? Skip and move on.
• Want to be specific? Type it, short, again.

When it's done, it assembles the prompt for you and starts rendering.

While the render generates

A render takes about 80 seconds, and you don't have to sit and wait in front of the screen: when it's done, the platform plays a subtle sound as a notification to let you know.

That time is yours. While one image generates, you can create other perspectives and keep launching renders, or edit another image, all while the rest renders in parallel. The flow never stops.

Generate a new version or Edit this image

The first render is almost never the final one, and that's fine. When it appears, you have two paths, and you almost always know which one without thinking:

• Generate a new version. For big changes. If the whole scene doesn't convince you (the surroundings, the mood, the level of realism, switching day to night), the AI remakes the image from scratch with the new cues.
• Edit this image. For targeted touch-ups. You like almost everything, but you want to remove an object, add a tree, or change a material: you mark that area with Select area, tell it what to do, and the rest stays intact. When you want to go deeper, see Refine and Iterate.

Quick rule: if more than about a third of the image changes, generate a new version; if it's a detail, edit. In two or three passes you'll have something to show.