Spline offers Spell, an experimental AI model that generates entire navigable 3D scenes from a single uploaded image or text prompt. Spell is available as a separate subscription at spell.spline.design and is currently in a limited early-access phase.

Spell is a diffusion model built by Spline that generates 3D worlds with multi-view consistency, supporting rendering via Gaussian Splatting. It accepts a text prompt or a single image as input and outputs a volumetric 3D scene that can be exported as a video, image sequence, or volumetric file, and embedded on the web. The model was trained on a proprietary dataset of real-world captures and licensed synthetic 3D objects; Spline states it did not use Spline user data for training.