Gnosika

Gnosika is a cinematic lab powered by AI, but rooted in something older.

[Gnosika @gnosikastudios]

We’re constantly running experiments as we work on anchoring AI-generated video inside stable, reusable 3D environments, and this is one of them.

[Gnosika @gnosikastudios]

In 1954, Joyce Collin-Smith travels to Rio Naza, Mexico, to visit Rodney Collin, the devoted pupil of the philosopher P.D. Ouspensky, who had died seven years earlier.

Full film available on our channel.

#CapCutCREAITE

[Gnosika @gnosikastudios]

"The Curtain of Time" is a short film by Amir Bozorgzadeh. In 1954, Joyce Collin-Smith visits Rodney Collin in Mexico. Pressing him about his teaching, she encounters the miraculous: his face begins to change before her eyes, revealing one life after another.

[Gnosika @gnosikastudios]

The next challenge for Gaussian splatting is not only reconstruction quality. It is interoperability. Not just producing a good-looking splat, but making captured worlds easier to move, review, version, and reuse across tools.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

While 3D Gaussian splatting represents a static scene as millions of semi-transparent 3D ellipsoids, each with position, scale, orientation, opacity, and view-dependent color, 4D Gaussian splatting extends that idea through time.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

A Gaussian splat flythrough is the easiest part to understand: capture a place, render it in real time, move through it. The deeper shift comes when that captured scene becomes addressable.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Most people see a Gaussian splat and think: impressive 3D render. But the render is only the visible output. A well-captured splat can preserve information that a flat image cannot: camera position, scale, geometry, confidence, and spatial context.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

One of the most interesting parts of Gaussian splatting is how it captures appearance. A splat doesn't just store color. It can store how appearance changes with viewing angle.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

A radiance field is a way of describing how light exists in a 3D space. Given a position and a viewing direction, it tells you what color and density should appear from that point of view.

[Gnosika @gnosikastudios]

Gaussian splatting doesn’t invent navigable 3D scenes. Photogrammetry, volumetric capture, and NeRFs were already moving in that direction. What feels new is the representation itself.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

The easiest way into Gaussian splatting is a molecule. A molecule isn't a picture of atoms, it's a structure of points sitting in space. You can picture it from any angle, drift around it, move between the atoms. Nothing about it is "framed."

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

With 4D Gaussian splats, a captured scene can become something closer to a place: navigable, viewable from different angles, and persistent enough to return to. You do not just watch the moment. You can move through it.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Most AI video clips last a few seconds, then the world disappears. That is one of the traps AI cinema is stuck in: generate a scene, the set is gone, generate the next, and you often get a new space, new geometry, and no reliable continuity.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Not a 360 photo, and not quite a traditional 3D model. More like a navigable reconstruction of a place, with depth, light, scale, and eventually time. The goal is to make real-world spaces easier to share, revisit, and build on.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

3D Gaussian splatting captures a scene as a set of volumetric points with position, shape, opacity, and view-dependent color. That means the scene can be rendered from new camera positions, and its appearance changes as the camera moves.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

3D Gaussian Splatting gave us real-time photorealistic worlds: millions of tiny 3D blobs carrying color, opacity, and view-dependent appearance.

Once it runs in real time, it stops being a scan and starts becoming a world.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Pixels are flat. Gaussians are not. Gauset captures reality as millions of soft volumetric Gaussian splats. Each splat carries position, color, and view-dependent appearance, so the scene can be rendered from new angles instead of locked to one frame.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Space has three dimensions. Time is the fourth. A photograph captures a viewpoint. A video captures a sequence. But reality is not a viewpoint or a sequence. It is spatial, temporal, and navigable.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Announcing Gauset. We’re building a platform for Gaussian spatial media, from capture to navigable scene. A moment becomes a place. A world you can return to, direct, and reuse.

→ https://t.co/s6BlMtRYod

[Gnosika @gnosikastudios]

Grateful to share that “Plastics” has received the Best Narrative Short award at the @aifilmfestorg.

Festival editor review:
"The film captures the mystery, surprise, and almost supernatural wonder of that encounter with Gurdjieff, weaving atmosphere and voice into a spell."