CubeComposer turns ordinary footage into full 360-degree panoramic video
Researchers just open-sourced CubeComposer, a video model that converts standard single-camera video into full 360-degree panoramic footage with 4K enhancement. By breaking frames into six cube faces and using sparse attention, it fills in what was behind the lens far better than tools like Argus, though minor distortion remains.
CubeComposer is an open-source video model that takes ordinary footage from a single camera and expands it into a full 360-degree panoramic video, complete with 4K enhancement.
How it reconstructs the blind spots
The model breaks the spherical environment into six cube faces using a diffusion framework. To keep transitions smooth and coherent across views, it relies on a sparse attention context pool that tracks temporal frames alongside adjacent cube faces.
That mechanism lets it hallucinate what was happening directly behind the camera lens. While there is still some noticeable distortion in the generated unseen areas, the overall visual quality is a clear step up from earlier alternatives like Argus. Details and demos are available on its project page at lg-li.github.io/project/cubec.
Why it matters
For creators building VR environments or spatial video assets, this offers a practical way to turn conventional archive footage into immersive scenes without shooting with a 360-degree rig. It outshines Argus on visual fidelity, though the slight warping means you will probably want to use it for background expansion or rapid prototyping rather than pristine hero shots.
Sources (2)
- lg-li.github.io(opens in a new tab)Originallg-li.github.io
- @aisearchio(opens in a new tab)Discussionx.com