Aligning Visual Foundation Encoders to Tokenizers for Diffusion Models
This work addresses the need for semantically rich tokenizers in diffusion models for image generation, offering a scalable solution with broad applications in AI-driven content creation.
The paper tackles the problem of aligning pretrained visual encoders as tokenizers for latent diffusion models in image generation, resulting in accelerated convergence (gFID of 1.90 within 64 epochs on ImageNet 256x256) and improved performance in text-to-image models.
In this work, we propose aligning pretrained visual encoders to serve as tokenizers for latent diffusion models in image generation. Unlike training a variational autoencoder (VAE) from scratch, which primarily emphasizes low-level details, our approach leverages the rich semantic structure of foundation encoders. We introduce a three-stage alignment strategy: (1) freeze the encoder and train an adapter and a decoder to establish a semantic latent space; (2) jointly optimize all components with an additional semantic preservation loss, enabling the encoder to capture perceptual details while retaining high-level semantics; and (3) refine the decoder for improved reconstruction quality. This alignment yields semantically rich image tokenizers that benefit diffusion models. On ImageNet 256$\times$256, our tokenizer accelerates the convergence of diffusion models, reaching a gFID of 1.90 within just 64 epochs, and improves generation both with and without classifier-free guidance. Scaling to LAION, a 2B-parameter text-to-image model trained with our tokenizer consistently outperforms FLUX VAE under the same training steps. Overall, our method is simple, scalable, and establishes a semantically grounded paradigm for continuous tokenizer design.