Skip to main content

Platform Concepts

How AI Processing Works

How AI Processing Works

MenuGaze uses an AI-powered pipeline to transform your food photos into polished menu-ready images and, for 3D Menu Pro, interactive 3D models. The entire process is automated — you upload a photo and the system handles the rest.

The Processing Pipeline

When you upload a photo of a menu item, it enters an asynchronous processing queue. The pipeline runs through the following stages:

  1. Background Removal — The AI identifies the dish in your photo and removes the background, isolating the food on a clean, transparent canvas. This ensures a consistent, professional look across all your menu items regardless of where you took the photo.
  2. AI Enhancement — Color correction, lighting adjustment, and sharpening are applied to make the dish look as appetizing as possible. The AI is trained on professional food photography to produce natural, vibrant results without over-processing.
  3. 3D Model Generation (3D Menu Pro only) — The enhanced photo is used as input for 3D model generation. The AI constructs a three-dimensional representation of the dish that customers can interact with in their browser.
  4. Model Optimization (3D Menu Pro only) — The generated 3D model passes through an optimization pipeline to ensure fast loading on mobile devices. The optimizer runs several stages: deduplication, pruning, welding, simplification, quantization, texture compression, and Draco compression. The final output is a GLB file capped at 2 MB.

Processing Times

For PhotoMenu, background removal and enhancement typically complete in under 30 seconds. For 3D Menu Pro, the full pipeline (including 3D generation and optimization) takes one to three minutes per item. Processing runs asynchronously in a background queue, so you can continue adding items while earlier ones are being processed.

Model Status Tracking

For 3D Menu Pro items, each model moves through a series of states that you can track from your dashboard:

  • PENDING — The item is queued for processing
  • GENERATING — The 3D model is being generated from your photo
  • PROCESSING — The model is being optimized for web delivery
  • READY — The model is live and visible to customers
  • FAILED — Processing encountered an error (you can retry from the dashboard)

If a model fails, you can retry generation from the dashboard. Common causes of failure include low-resolution photos, images with cluttered backgrounds, or photos where the dish is partially obscured.