PixSDS: Why Latent SDS Makes Noisy Pixels

• URL: http://arxiv.org/abs/2608.12997v1
• Date: Thu, 13 Aug 2026 09:20:58 GMT
• Title: PixSDS: Why Latent SDS Makes Noisy Pixels
• Abstract summary: Score Distillation Sampling (SDS) enables text-to-3D generation by optimizing rendered images with a pretrained diffusion prior.<n>We propose PixSDS, a lightweight VAE-consistent gradient repair method.<n> Experiments in 2D optimization and text-to-3D generation show that PixSDS substantially reduces structured artifacts while preserving semantic content.
• Score: 0.0
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: 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/.

Related papers

• DST-Net: A Dual-Stream Transformer with Illumination-Independent Feature Guidance and Multi-Scale Spatial Convolution for Low-Light Image Enhancement [9.193933558950341]
  Low-light image enhancement aims to restore the visibility of images captured by visual sensors in dim environments.<n>We propose a Dual-Stream Transformer Network (DST-Net) based on illumination-agnostic signal prior guidance and multi-scale spatial convolutions.<n>Our method achieves a PSNR of 25.64 dB on the LOL dataset.
  arXiv  Detail & Related papers  (2026-03-17T13:07:32Z)
• DSA-SRGS: Super-Resolution Gaussian Splatting for Dynamic Sparse-View DSA Reconstruction [67.42242016220122]
  Digital subtraction angiography is a key imaging technique for the auxiliary diagnosis and treatment of cerebrovascular diseases.<n>Recent advancements in gaussian splatting and dynamic neural representations have enabled robust 3D vessel reconstruction from sparse dynamic inputs.<n>This paper proposes DSA-SRGS, the first super-resolution gaussian splatting framework for dynamic sparse-view DSA reconstruction.
  arXiv  Detail & Related papers  (2026-03-05T03:41:08Z)
• Decompositional Neural Scene Reconstruction with Generative Diffusion Prior [64.71091831762214]
  Decompositional reconstruction of 3D scenes, with complete shapes and detailed texture, is intriguing for downstream applications.<n>Recent approaches incorporate semantic or geometric regularization to address this issue, but they suffer significant degradation in underconstrained areas.<n>We propose DP-Recon, which employs diffusion priors in the form of Score Distillation Sampling (SDS) to optimize the neural representation of each individual object under novel views.
  arXiv  Detail & Related papers  (2025-03-19T02:11:31Z)
• RewardSDS: Aligning Score Distillation via Reward-Weighted Sampling [14.725841457150414]
  RewardSDS weights noise samples based on alignment scores from a reward model, producing a weighted SDS loss.<n>This loss prioritizes gradients from noise samples that yield aligned high-reward output.<n>We evaluate RewardSDS and RewardVSD on text-to-image, 2D editing, and text-to-3D generation tasks.
  arXiv  Detail & Related papers  (2025-03-12T17:59:47Z)
• Consistent3D: Towards Consistent High-Fidelity Text-to-3D Generation with Deterministic Sampling Prior [87.55592645191122]
  Score distillation sampling (SDS) and its variants have greatly boosted the development of text-to-3D generation, but are vulnerable to geometry collapse and poor textures yet. We propose a novel and effective "Consistent3D" method that explores the ODE deterministic sampling prior for text-to-3D generation. Experimental results show the efficacy of our Consistent3D in generating high-fidelity and diverse 3D objects and large-scale scenes.
  arXiv  Detail & Related papers  (2024-01-17T08:32:07Z)
• StableDreamer: Taming Noisy Score Distillation Sampling for Text-to-3D [88.66678730537777]
  We present StableDreamer, a methodology incorporating three advances. First, we formalize the equivalence of the SDS generative prior and a simple supervised L2 reconstruction loss. Second, our analysis shows that while image-space diffusion contributes to geometric precision, latent-space diffusion is crucial for vivid color rendition.
  arXiv  Detail & Related papers  (2023-12-02T02:27:58Z)
• Contrastive Denoising Score for Text-guided Latent Diffusion Image Editing [58.48890547818074]
  We present a powerful modification of Contrastive Denoising Score (CUT) for latent diffusion models (LDM) Our approach enables zero-shot imageto-image translation and neural field (NeRF) editing, achieving structural correspondence between the input and output.
  arXiv  Detail & Related papers  (2023-11-30T15:06:10Z)
• Pixel-Aware Stable Diffusion for Realistic Image Super-resolution and Personalized Stylization [23.723573179119228]
  We propose a pixel-aware stable diffusion (PASD) network to achieve robust Real-ISR and personalized image stylization. A pixel-aware cross attention module is introduced to enable diffusion models perceiving image local structures in pixel-wise level. An adjustable noise schedule is introduced to further improve the image restoration results.
  arXiv  Detail & Related papers  (2023-08-28T10:15:57Z)
• Delta Denoising Score [51.98288453616375]
  We introduce Delta Denoising Score (DDS), a novel scoring function for text-based image editing. It guides minimal modifications of an input image towards the content described in a target prompt.
  arXiv  Detail & Related papers  (2023-04-14T12:22:41Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.