Learning to Watermark in the Latent Space of Generative Models

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.

As AI-generated images become indistinguishable from reality, invisible watermarks are crucial to verify content origins and prevent malicious misuse like deepfakes. However, traditional watermarking techniques are applied after an image is fully created, which allows users to easily bypass the security feature by removing a single line of code. To solve this, we developed DistSeal, a framework that embeds the watermarking step directly in the generation process. By distilling the watermarking model into the generation model, the watermark becomes an inseparable, automatic part of the image generation process. This approach stops users from simply disabling the watermarking code in open-source releases while also removing the extra time overhead on top of generation introduced by watermarking.