Interactive imagery for places and people
People shouldn’t have to search through pages to picture your place. Here they ask, and hear the answer in seconds while the picture moves to what it’s about.
The previews use fixed questions and recorded answers. Arrange a live walkthrough to try asking your own questions by voice.
An invented propertyORIELLAA place you can explore.Ask about the place. See what unfolds.
Built by ThinkSpatial to show the system · Fixed questions, recorded answers, no account. Oriella is invented: there is no such hotel, and this is not a client’s website.
An invented architectMAYA FERREIRAA person you can ask.Ask what you would ask on a first call.
Built by ThinkSpatial to show the system · Fixed questions, recorded answers, no account. Maya Ferreira is invented: she does not exist, she did not make this, and it is not her website.
Two previews, one system
Both work the same way: a place or a person you explore through conversation. For a person, we call it a Living Portrait. These public previews use fixed questions and curated, synthesized voice tracks, with answers reviewed before recording. Open either in the browser with no account.
New · a second way into your website
Most websites have a button in the corner that opens a chat box. This one opens the place. Visitors choose how to take it in: move through the pictures, talk to a guide, or type a question and get your own approved answer. Your website keeps its pages, forms and booking. This sits on top of them, with no rebuild.
It opens Oriella, an invented hotel built to show the system. Nothing in its chat is written by a model: every answer is one the owner approved.
One system, different expressions
Both previews above are the third of these. The same system also answers across an existing site, or inside a single image. Each one starts with the information behind your project.
Ask a question. See what it points to.
In a live deployment, visitors ask their own questions and follow up by voice. The guide answers from your approved information and brings the relevant place or detail into view. Try this in an arranged live walkthrough; the public previews use fixed questions and recorded answers.
An image becomes a place to explore.
Start with a photograph or rendering. ThinkSpatial identifies the objects and places inside it and connects them to your approved information, so visitors can explore a room, zoom into a detail and discover what they’re looking at.
A conversation that moves the scene.
Let the guide bring scenes, imagery and video into view as the conversation develops. An experience fitted to the website and the visitor’s screen.
We build it
Send us your website. We’ll make a preview of your place from the photos and details already on it, within two working days, and send you a private link. If it’s useful, we’ll talk about putting it on your site.
Nothing to prepare, and nothing to pay to see it.
Ask for a previewThe ThinkSpatial system
The source is your documented repository: architectural briefs, hospitality dossiers, archival catalogs, photographs and transcripts. CAD-derived plans and renderings can join as exported documents and imagery; native CAD integration is scoped separately.
Structured facts and source references give the guide context for what to say and what to show. Ingestion starts from existing records, with review to resolve gaps and approve content before publication.
Engine Architecture
Repeatable build pipelines turn approved inputs into a connected experience. Model-assisted extraction and editorial review prepare the material; explicit bindings make published question routing and recorded playback deterministic.
Ingest raw assets, documents and transcripts from a supplied client repository. Organize approved information into fact schemas with stable identifiers, source references and the questions each fact can answer.
Source material → reviewed fact records
Vision-assisted extraction locates image regions as normalized coordinates and outlines. The build binds these regions to the node graph, assigns facts to scenes and resolves question destinations to stable scene identifiers.
Node graph → image coordinates → scene routes
Reviewed answers are synthesized and bound to question identifiers and source fingerprints. Public previews play stored audio and follow fixed routes. Live deployments use configurable voice providers, including OpenAI, ElevenLabs and Cartesia, with streaming paths to reduce response delay.
Audio binding → recorded playback or live voice
Place tours and portraits have distinct build pipelines and share the published viewer and recorded-answer runtime. New deployments supply their own records, brand and scene configuration. Content approval remains part of production; live responses and generated imagery are not deterministic.
Structured case breakdown / demonstration
Maya is an invented architect, not a client or an endorsement. Her Living Portrait demonstrates how a documented practice becomes a navigable spatial narrative through the portrait build pipeline.
Build snapshot: 22 September 2026. Scenes, fact records and question routes are separate counts.
Explore the Maya demonstration