Fast and High-Resolution Latent Decoding
with Pixel Diffusion
TL;DR: PiD directly decodes latent representations into high-resolution images, replacing the decode–then–super-resolve cascade while achieving lower latency and higher visual quality. *
Abstract #
Most practical high-resolution text-to-image systems rely on latent diffusion models, where generation is performed in a compact latent space and a decoder maps latents back to pixels. Yet the latent-to-pixel decoder is reconstruction-oriented, optimized to invert the encoder rather than synthesize more details, and becomes increasingly costly at megapixel scale. This drawback calls for a more expressive and efficient decoding paradigm. Motivated by recent progress in scalable pixel-space diffusion, we introduce PiD, a Pi xel diffusion D ecoder that reformulates latent decoding as conditional pixel diffusion, unifying decoding and upsampling into one generative module. By denoising directly in high-resolution pixel space, PiD synthesizes 4× and even 8× upscaled images with low latency. For latent conditioning, a lightweight sigma-aware adapter injects noise-corrupted latents into the pixel diffusion backbone, enabling PiD to decode partially denoised latents and terminate the latent diffusion process early. To further improve efficiency, we distill the model using DMD2, reducing inference to just 4 steps. PiD applies to both conventional VAE latents and semantic latents (e.g., SigLIP, DINOv2) used in recent RAE-based models. PiD decodes latents of 512×512 images into 2048×2048 pixels in under 1 second with 13 GB peak memory on a consumer RTX 5090, and as fast as 210 ms on a GB200 GPU, about 6× faster than cascaded diffusion-based super-resolution pipelines with better visual fidelity.
Results #
From Latent to Pixels
4K Decode
Baseline Comparison
Quantitative Results (Decoding + Upsampling, 512² → 2048²)
PiD is up to 5.9× faster than SeedVR2 (211.2 ms vs 1237.5 ms)
% of evaluations where judges prefer PiD over each baseline