Learning to Watermark in the Latent Space of Generative Models

TL;DR

This paper introduces DistSeal, a novel latent space watermarking method for generative models that is more efficient and robust than pixel-space approaches, enabling faster and imperceptible watermarks.

Contribution

The work presents a unified latent watermarking approach applicable to diffusion and autoregressive models, with effective distillation into models for in-model watermarking.

Findings

Latent watermarks are up to 20x faster than pixel-space baselines.

Distilled latent watermarkers outperform pixel-space watermarkers in robustness.

The approach maintains imperceptibility while improving efficiency.

Abstract

Existing approaches for watermarking AI-generated images often rely on post-hoc methods applied in pixel space, introducing computational overhead and potential visual artifacts. In this work, we explore latent space watermarking and introduce DistSeal, a unified approach for latent watermarking that works across both diffusion and autoregressive models. Our approach works by training post-hoc watermarking models in the latent space of generative models. We demonstrate that these latent watermarkers can be effectively distilled either into the generative model itself or into the latent decoder, enabling in-model watermarking. The resulting latent watermarks achieve competitive robustness while offering similar imperceptibility and up to 20x speedup compared to pixel-space baselines. Our experiments further reveal that distilling latent watermarkers outperforms distilling pixel-space ones, providing a solution that is both more efficient and more robust.