Upload a reference
Start from a product, prop, character object, scan, or concept image.
Image to 3D object studio
Turn one image into a web-viewable 3D object.
ML-LiTo is a public front door for experimenting with Apple LiTo-style image-to-3D generation: upload a product, prop, collectible, or object reference and generate a browser-viewable Gaussian splat through an on-demand hosted GPU.
single image input
surface light field tokens
web-viewable 3D
Create from one image
Bring a clean object image.
Upload one centered object image. The hosted LiTo pipeline isolates the subject, samples the 3D representation, and returns an interactive Gaussian splat you can view, share, or download.
Hosted workflow
A practical LiTo pipeline for creators and sellers.
Generate 3D
Run the open LiTo image-to-3D model on a hosted GPU or Apple Silicon worker.
View the splat
Open the result in a browser viewer, inspect the object, and share the generated 3D preview.
Export or refine
Download the Gaussian splat or continue cleanup and publishing in SuperSplat.
Research base
Built around Apple’s public LiTo release.
Apple describes LiTo as a latent 3D representation for jointly modeling geometry and view-dependent appearance, including effects like reflections and lighting. The public repository includes an interactive image-to-3D demo, pretrained checkpoints, and Apple Silicon support for the demo path.
ML-LiTo on Almonte Technology is independent and is not an Apple product. It is being built around Apple’s public `apple/ml-lito` research code and model release.