Hi there, is there any way to create 3D models from 2D images? Essentially photogrammetry just short of scanning an object with lidar? I’ve tried a few different (non-Google) models which have subpar outputs and I’m hoping that with Nano Banana, the outputs can be photorealistic, true to color and true to texture with respect to 3D model generation. Thanks!
Hi @Nikhil_Bafna
Welcome to the Google AI Forum!!!
Thank you for reaching out!
Could you please share some more details like, what kind 3D model you want to create?
Hey,
Hope you’re keeping well.
Google Cloud doesn’t have a dedicated photogrammetry service, but you can use Vertex AI with custom or fine-tuned 3D generation models to process 2D images into 3D meshes or textured models. The “Nano Banana” model in AI Studio is designed for image generation, so for true-to-texture 3D output you’d need to integrate it with an external 3D reconstruction pipeline, then host and run the workflow in Vertex AI Workbench or via custom prediction endpoints. You can store intermediate assets in Cloud Storage and use Cloud Functions or Cloud Run to automate the conversion steps. If you need photorealistic textures, consider combining the generated geometry with texture maps created by a diffusion model in Vertex AI.
Thanks and regards,
Taz