What happens when you combine Nano Banana with Gaussian splat tours?

Hey everyone, I’ve been experimenting with combining **Nano Banana-style image reimagining** with **Gaussian splat tours**. Gaussian splats are amazing for experiencing a real space as it exists. Nano Banana is great at reimagining how an image could look. So the question was: **Why keep them apart?** We built [Spatial Studio](https://realhorizons.ai/) (by realhorizons), we built a feature called **Reframe AI** that lets users move through a 3D splat tour, pick a camera view, and reimagine that exact perspective using AI. Some examples: * An empty room can be furnished. * A cafe can be changed into a seasonal theme. * An under-construction property can be visualized closer to completion. * A venue can preview different event setups. * A real estate team can create multiple marketing concepts from one capture. We’re also experimenting with **Spatial Props**, where selected AI-generated objects or design elements can be brought back into the **3D experience** instead of staying as a flat edited image. The flow is: **Capture space → generate splat tour → pick a view → reimagine with AI → publish/share** So the original captured splat stays intact, but AI-generated possibilities can be layered into the spatial tour. Still early, but I think this combo is interesting: **real-world 3D capture + Gaussian splats + Nano Banana-style image generation = reimagined spatial experiences**