DreamCache: Finetuning-Free Lightweight Personalized Image Generation via Feature Caching

TL;DR

DreamCache is a scalable, efficient method for personalized image generation that uses feature caching and lightweight adapters to achieve high-quality results with fewer parameters and lower computational costs.

Contribution

It introduces a novel feature caching technique combined with trained adapters, enabling high-quality personalized image generation without extensive retraining.

Findings

Achieves state-of-the-art image and text alignment.

Uses an order of magnitude fewer parameters than existing methods.

Offers more computationally efficient and versatile personalized image generation.

Abstract

Personalized image generation requires text-to-image generative models that capture the core features of a reference subject to allow for controlled generation across different contexts. Existing methods face challenges due to complex training requirements, high inference costs, limited flexibility, or a combination of these issues. In this paper, we introduce DreamCache, a scalable approach for efficient and high-quality personalized image generation. By caching a small number of reference image features from a subset of layers and a single timestep of the pretrained diffusion denoiser, DreamCache enables dynamic modulation of the generated image features through lightweight, trained conditioning adapters. DreamCache achieves state-of-the-art image and text alignment, utilizing an order of magnitude fewer extra parameters, and is both more computationally effective and versatile than existing models.