A world that
stays itself.
Walk around a corner. Look back. What you discovered should still be there. What you changed should stay changed.
An independent exploration
What if you could step inside an idea?
Explore it. Change it. Understand it together.
84 seconds / A research film
The places, the progress, and the question behind Sciescape.
Original panoramas and recorded prototype outputs. A complete live journey remains the research goal.
Open the filmFive generated places introduce the ambition: a world that stays coherent as you move and change it, with a lifelike guide grounded in what you see. The film separates existing visual outputs from that goal. It does not demonstrate a complete live conversation or freely walkable world.
Read the timed transcript · Film credits & provenance01 — The beginnings
Five places imagined by the Sciescape pipeline.
Choose one. Take a look around.
Follow the quiet.
Stay for the blue hour.
Let the outside wait.
A place to get lost in thought.
Find a different kind of time.
These are image-based previews. Free movement, live conversation and persistent edits are the research ahead. Behind the images

02 — The ambition
We’re working toward a new way to explore an idea:
a world you can talk into being, with someone to help you understand it.
Walk around a corner. Look back. What you discovered should still be there. What you changed should stay changed.
A lifelike presence that follows your attention, reasons with you, and makes the unfamiliar understandable.
“What if…?” becomes something to explore. Speak naturally, interrupt, ask again. Let curiosity lead.
03 — The work behind the wonder
Can what you see, where you move, what you change, and what the guide knows all belong to one persistent world?
That is the question driving Sciescape. The goal is a dependable stack for creating and exploring worlds through conversation—and evidence that explains why it works.
Read the research directionAn honest view of where we are
Five real panoramic outputs, preserved with their original files and provenance.
Geometry, movement and edits that agree across viewpoints. Reliability remains an open problem.
A stronger 3D avatar, grounded reasoning and natural voice, working as one continuous experience.
An open invitation
Interested in helping build it?