method · first seen 5 Oct 2026
pixel-space diffusion
Also: pixel-space flow models, pixel-space diffusion model, pixel-space video diffusion
Diffusion image generation operating directly on pixels rather than in a latent autoencoder space.
Pixel-space diffusion trains image and video generators directly on raw pixels instead of inside a pretrained VAE latent space. That avoids lossy compression and two-stage training. What is new in these sources is that the approach is being made to work at scale. Iris-3B is a 3B open text-to-image model with weights and an Apple Silicon port. A two-person lab released a 2B Pyramid-JiT model trained for about $21K and posted it on Show HN. Several papers study why v-prediction fails, add adaptive and multi-scale tokenization, or convert pretrained latent video models (PixVid, JVT, U-JiT) into pixel space.
If pixel-space models match latent models in quality and cost, the VAE could drop out of the generation pipeline. That would allow end-to-end training, finer detail for restoration and depth tasks, and per-frame control in video. Several independent groups are publishing open weights and papers, which suggests this is a real research direction and not one vendor's push.
Analyst note (emerging), 17 Nov 2026.
Mentions per week
PapersCodeDiscussion
Rank 6 on 16 Nov 2026 · 13 mentions in 28 days · 13 independent groups · code → papers → discussion · first seen 6 weeks ago · acceleration +0.88
Earliest sources
- speridlabs/iris-3bCode · 5 Oct 2026 · speridlabs · 229 stars
- Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-TuningPapers · 7 Oct 2026 · Hanqiu Li Cai, Chema Garabito
- BudgetPix: Compute-Adaptive Tokenization for Pixel-Space Image DiffusionPapers · 8 Oct 2026 · Ozgur Kara, Yujia Chen, Daniel Watson, David Forsyth, James Matthew Rehg, Wen-Sh…
- TKY-27/iris-3b-metalCode · 9 Oct 2026 · TKY-27 · 58 stars
- PixelMoE: Scaling Pixel-Space Diffusion with LLM-Style Mixture-of-ExpertsPapers · 13 Oct 2026 · Ruixiao Dong, Jianmin Bao, Bozhi Luan, Zijie Meng, Xin Tao, Pengfei Wan, Kun Gai…
- Rethinking the design space of Vanilla Pixel Diffusion TransformersPapers · 16 Oct 2026 · Soumik Mukhopadhyay, Long Mai, Abhinav Shrivastava, Aniruddha Mahapatra
Latest sources
- ZhouTimeMachine/JVTCode · 16 Nov 2026 · ZhouTimeMachine · 39 stars
- JVT: Just Video Transformer makes pixel-space video diffusion simplePapers · 13 Nov 2026 · Jianjun Zhou, Hua Geng, Haoyu Guo
- Back to Pixels: Lifting Latent Video Diffusion into Pixel SpacePapers · 12 Nov 2026 · Yanke Song, Shuo Chen, Chris Rockwell, Guande He, Nan Liu, Desai Xie, Tianpei Gu…
- What kills $v$-prediction? A Patch-wise PCA Perspective on Pixel-Space Flow MatchingPapers · 9 Nov 2026 · Tianze Luo, Haofeng Huang, Yuan Yao
- Polynomial Denoisers on the Image Manifold: Volterra Networks for Self-Contained Pixel DiffusionPapers · 7 Nov 2026 · Siddharth Roheda, Rohit Chowdhury, Amit Satish Unde, Aniruddha Bala
- Scale as a Modality: Multi-Scale Pixel-Space Diffusion TransformersPapers · 5 Nov 2026 · Ruibo Chen, Kaishen Wang, Tianyi Xiong, Chenxi Liu, Yanshuo Chen, Heng Huang, Ji…
- Show HN: A decent text-to-image model trained for ~$20K (Apache 2.0 Weights)Discussion · 4 Nov 2026 · schopra909 · 287 points
- Linum-AI/pjit-21kCode · 3 Nov 2026 · Linum-AI · 80 stars
- Pixel-space video generation with U-JiTPapers · 3 Nov 2026 · Théophane Vallaeys, Jakob Verbeek, Matthieu Cord
- THUMAI-Lab/RiTCode · 1 Nov 2026 · THUMAI-Lab · 58 stars
- Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow ModelsPapers · 26 Oct 2026 · Ziqiao Wang, Yang You
- HiFiPix: Bringing High-Fidelity Real-World Super-Resolution Back to Pixel SpacePapers · 26 Oct 2026 · Wenjie Li, Du Chen, Hui Wang, Linlong Fan, Youwei Lv, Zeyi Chen, Zehan Chen, Jin…
Mentioned together with
In the ledger
- 2026-W47 · call #9 · open