IMAGE-ALCHEMY: Advancing subject fidelity in personalised text-to-image generation

TL;DR

This paper introduces IMAGE-ALCHEMY, a novel two-stage method for personalized text-to-image generation that maintains subject fidelity and broad scene diversity by fine-tuning attention weights with LoRA on SDXL.

Contribution

The paper presents a new pipeline combining LoRA fine-tuning and segmentation-driven Img2Img to improve personalized subject fidelity without sacrificing scene diversity.

Findings

Achieves a DINO similarity score of 0.789 on SDXL.

Outperforms existing personalized text-to-image methods.

Preserves broader generative capabilities while integrating new subjects.

Abstract

Recent advances in text-to-image diffusion models, particularly Stable Diffusion, have enabled the generation of highly detailed and semantically rich images. However, personalizing these models to represent novel subjects based on a few reference images remains challenging. This often leads to catastrophic forgetting, overfitting, or large computational overhead.We propose a two-stage pipeline that addresses these limitations by leveraging LoRA-based fine-tuning on the attention weights within the U-Net of the Stable Diffusion XL (SDXL) model. First, we use the unmodified SDXL to generate a generic scene by replacing the subject with its class label. Then, we selectively insert the personalized subject through a segmentation-driven image-to-image (Img2Img) pipeline that uses the trained LoRA weights.This framework isolates the subject encoding from the overall composition, thus preserving SDXL's broader generative capabilities while integrating the new subject in a high-fidelity manner. Our method achieves a DINO similarity score of 0.789 on SDXL, outperforming existing personalized text-to-image approaches.