Visualizations let a website visitor upload a photo of their project and see an AI-rendered preview of the requested work applied to it. The widget assistant generates the image inside the chat, turning a casual question into a visual that helps convert visitors into leads.

The flow is conversational: the AI drives it, and the visitor just answers. When a visitor describes a potential project, the assistant shows an upload button so they can share a photo of their site. Once a photo is uploaded and the visitor has described the services they want, the assistant renders the preview.

While the image generates, the chat shows a Generating preview placeholder. The finished render appears as an image in the conversation, ready for the visitor to react to or refine.

The assistant only generates a visualization after a photo has actually been uploaded in the conversation, and it confirms which services to include before the first render.

A visitor can ask to change a finished preview, for example "add another element" or "remove that detail." When they do, the assistant edits the last generated image rather than starting over from the original photo. This keeps earlier changes intact while layering on the new request.

To carry edits forward correctly, the assistant includes every desired element in the new request, not just the latest change.

The Services you configure on the widget define what visitors can ask to see rendered. See widget setup for how to add and order services.

Your organization's trade also shapes the output. The trade sets the rendering style and vocabulary. A landscaping business produces photorealistic yard renderings with golden-hour lighting, while a painting business produces interior renderings focused on paint colors and finishes. Visitors get results tuned to your line of work without any extra configuration.

Each visualization uses your company's AI usage. The widget checks before it makes an image, so a visitor can't run up image costs when your AI usage is used up.

If this month's AI usage is used up, the preview won't be made and the assistant tells the visitor it can't produce the image. Visualizations come back when your usage refills or an admin turns on extra usage. See How AI usage works to learn more.

Encourage visitors to upload clear, well-lit photos taken straight on. The sharper the source photo, the more accurate the rendering. Visitors who name the specific services they want (rather than asking for vague "improvements") get previews that match their intent more closely.