PixSDS: Why Latent SDS Makes Noisy Pixels

Hugging Face Daily Papers Papers
score-distillation-sampling text-to-3d vae latent-sds diffusion-models 3d-generation artifacts

Summary

PixSDS identifies VAE-induced pixel drift in latent score distillation sampling and proposes a gradient repair method that decodes latent SDS lookahead steps to guide pixel-space optimization, reducing artifacts in text-to-3D generation.

Score Distillation Sampling (SDS) enables text-to-3D generation by optimizing rendered images with a pretrained diffusion prior, but latent SDS often produces structured color artifacts and high-frequency texture noise. We identify a failure mode of latent SDS caused by VAE-induced pixel drift: the optimized image can move along pixel-space directions that are weakly constrained by the VAE encoder, so its latent representation remains clean and semantically meaningful while the image itself accumulates visible artifacts. We support this diagnosis with controlled 2D SDS experiments, VAE-only optimization, and a simplified analysis showing that encoder-like latent objectives can amplify image-space noise when the inverse mapping to pixels is underconstrained. Motivated by this observation, we propose PixSDS, a lightweight VAE-consistent gradient repair method. PixSDS decodes a latent SDS lookahead step and uses the decoded image as a clean direction for pixel-space optimization, reducing motion in VAE-inconsistent directions without retraining the diffusion model, changing the renderer, or replacing the SDS objective. Experiments in 2D optimization and text-to-3D generation show that PixSDS substantially reduces structured artifacts while preserving semantic content. Code is publicly available at https://sevashasla.github.io/pixsds-webpage/.
Original Article
View Cached Full Text

Cached at: 08/14/26, 11:27 AM

Paper page - PixSDS: Why Latent SDS Makes Noisy Pixels

Source: https://huggingface.co/papers/2608.12997

Abstract

PixSDS fixes VAE-induced pixel drift in latent score distillation sampling by guiding optimization with decoded image directions, reducing artifacts in text-to-3D generation.

Score Distillation Sampling(SDS) enablestext-to-3D generationby optimizing rendered images with a pretrained diffusion prior, butlatent SDSoften produces structured color artifacts and high-frequency texture noise. We identify a failure mode oflatent SDScaused byVAE-induced pixel drift: the optimized image can move along pixel-space directions that are weakly constrained by theVAE encoder, so its latent representation remains clean and semantically meaningful while the image itself accumulates visible artifacts. We support this diagnosis with controlled 2D SDS experiments, VAE-only optimization, and a simplified analysis showing that encoder-like latent objectives can amplify image-space noise when the inverse mapping to pixels is underconstrained. Motivated by this observation, we proposePixSDS, a lightweight VAE-consistentgradient repairmethod.PixSDSdecodes alatent SDSlookahead step and uses the decoded image as a clean direction for pixel-space optimization, reducing motion in VAE-inconsistent directions without retraining the diffusion model, changing the renderer, or replacing the SDS objective. Experiments in 2D optimization andtext-to-3D generationshow thatPixSDSsubstantially reduces structured artifacts while preserving semantic content. Code is publicly available at https://sevashasla.github.io/pixsds-webpage/.

View arXiv pageView PDFProject pageGitHub0Add to collection

Get this paper in your agent:

hf papers read 2608\.12997

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.12997 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.12997 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.12997 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

PiD: Fast and High-Resolution Latent Decoding with Pixel Diffusion

Hugging Face Daily Papers

PiD introduces a pixel diffusion decoder that reformulates latent decoding as conditional pixel diffusion, enabling fast and high-quality image synthesis at high resolutions with reduced computational requirements. It decodes latents into 4x or 8x upscaled images in under a second on consumer hardware.

@FeitengLi: NVIDIA Spatial Intelligence Lab proposes PiD, redesigning the decoding stage in latent diffusion models. Current mainstream text-to-image generation happens in latent space, then uses a VAE decoder to map back to pixels. This decoder's…

X AI KOLs Timeline

NVIDIA Spatial Intelligence Lab proposes PiD, which redesigns the decoding stage of latent diffusion models as a conditional pixel diffusion process, unifying decoding and upsampling to achieve low-latency, high-resolution decoding.

L2P: Unlocking Latent Potential for Pixel Generation

Hugging Face Daily Papers

The L2P paper introduces a Latent-to-Pixel transfer paradigm that leverages pre-trained latent diffusion models to create efficient pixel-space models capable of 4K generation with minimal training overhead.

nvidia/PiD

Hugging Face Models Trending

NVIDIA releases PiD (Pixel Diffusion Decoder), a conditional pixel-space diffusion model that unifies latent-to-pixel decoding and upsampling into one generative module, producing super-resolved images in one pass. Model checkpoints and VAE weights are released under a non-commercial license.