Training-Free Text-to-Image Compositional Food Generation via Prompt Grafting

TL;DR

This paper introduces Prompt Grafting, a training-free method that improves multi-food image generation by controlling object separation using spatial cues, enhancing accuracy for dietary and recipe visualization applications.

Contribution

It proposes a novel training-free framework that combines explicit spatial prompts with implicit layout guidance to control food object separation in generated images.

Findings

Significantly improves target object presence in generated images

Enables controllable separation or mixing of food items

Demonstrates effectiveness across two food datasets

Abstract

Real-world meal images often contain multiple food items, making reliable compositional food image generation important for applications such as image-based dietary assessment, where multi-food data augmentation is needed, and recipe visualization. However, modern text-to-image diffusion models struggle to generate accurate multi-food images due to object entanglement, where adjacent foods (e.g., rice and soup) fuse together because many foods do not have clear boundaries. To address this challenge, we introduce Prompt Grafting (PG), a training-free framework that combines explicit spatial cues in text with implicit layout guidance during sampling. PG runs a two-stage process where a layout prompt first establishes distinct regions and the target prompt is grafted once layout formation stabilizes. The framework enables food entanglement control: users can specify which food items should remain separated or be intentionally mixed by editing the arrangement of layouts. Across two food datasets, our method significantly improves the presence of target objects and provides qualitative evidence of controllable separation.