MagicSpin 360° · Hackathon Build
A 360° view from a single photo.
Winner, Google Nano Banana Hackathon 2025. Turns a single 2D image into an interactive 360° rotation through a multi-model generation pipeline.
Single image to interactive rotation
Upload, AI generation, drawing, scene customization, frame synthesis, and a draggable 360° viewer.
- Result
- Winner, Nano Banana 2025
- Input
- A single 2D image
- Output
- Interactive 360° viewer
- Models
- Gemini · Stability · SAM
The pipeline.
Subject isolation
Segment Anything separates the subject from its background, so only the object rotates.
Scene understanding
Gemini Pro Vision reads the subject and infers the occluded geometry the back views will need.
Frame synthesis
Stability AI renders the rotation frames from the isolated subject and inferred context.
Spin
Sequential frames feed a draggable 360° viewer, with optional smoothing between frames.
Constraints.
Dedicated 360° capture needs hardware, arrays of cameras, or hours of manual editing. The brief allowed one input image and a weekend, so every stage uses a hosted model. Most of the engineering went into the hand-offs between them: what each model receives, what it is allowed to infer, and where a user can redraw the scene before frames are generated.