Breaking

NVIDIA Unveils Lyra 2.0, Which Generates Explorable 3D Worlds From a Single Image

November 6, 2026 at 15:28 EST

Media Generation
Research & Papers

NVIDIA has introduced Lyra 2.0, a research system that turns a single image into a 3D environment users can navigate along a camera path they choose.

NVIDIA · AI Research

One image in, a navigable 3D world out

Lyra 2.0 is a research system that turns a single image into a 3D environment. Users choose the camera path and explore the scene from viewpoints of their own selection.

How the system keeps a scene consistent

Generating views one after another lets small errors build up. The system reuses earlier views and learns to correct drift as the camera moves away from the source image.

STEP 1

Single input image

The starting point is one ordinary picture.

→

STEP 2

Generate along the path

Each new view draws on earlier views.

→

STEP 3

Correct accumulated error

Learned correction limits visual drift.

→

OUTPUT

Navigable 3D environment

Explorable from user-selected viewpoints.

Still to be verified

Lyra 2.0's performance and limits are not yet established.

Independent benchmarks

None were included in the initial release.

Long camera paths

How well worlds hold together over longer trajectories.

Comparison with existing methods

Head-to-head results against current 3D generation approaches.

Availability and outside review

Public access details and expert assessments are pending.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.