Native Multimodal Models are World Learners

HunyuanImage-3.0: A Powerful Native Multimodal Model for Image Generation

Fast and memory-efficient exact attention

[ICCV2025]Code Release of Harmonizing Visual Representations for Unified Multimodal Understanding and Generation

Voyager is an interactive RGBD video generation model conditioned on camera input, and supports real-time 3D reconstruction.

Official codebase for "Self Forcing: Bridging Training and Inference in Autoregressive Video Diffusion" (NeurIPS 2025 Spotlight)

PixiEditor is a Universal Editor for all your 2D needs

Generative Models by Stability AI

Official PyTorch Implementation of "Latent Denoising Makes Good Visual Tokenizers"

[ICCV2025] "Di[M]O: Distilling Masked Diffusion Models into One-step Generator", Yuanzhi Zhu, Xi Wang, Stéphane Lathuilière, Vicky Kalogeiton

This repo provides a working re-implementation of Latent Adversarial Diffusion Distillation by AMD

MoDM is a cache-aware, hybrid serving system that accelerates image generation by dynamically combining small and large diffusion models for efficient, high-quality output.

The open-source CapCut alternative

Locality-aware Parallel Decoding for Efficient Autoregressive Image Generation

[CVPR 2025 Highlight] VideoScene: Distilling Video Diffusion Model to Generate 3D Scenes in One Step

[ICCV2025]LazyMAR: Accelerating Masked Autoregressive Models via Feature Caching

Official Implementation of Diffusion Step Annealing (DiSA) in Autoregressive Image Generation

Janus-Series: Unified Multimodal Understanding and Generation Models

Official PyTorch implementation of the paper "dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching" (dLLM-Cache).

Official PyTorch implementation for "Effective and Efficient Masked Image Generation Models"

The official repo of continuous speculative decoding

12 Lessons to Get Started Building AI Agents

[NeurIPS 2024] ENAT: Rethinking Spatial-temporal Interactions in Token-based Image Synthesis

HART: Efficient Visual Generation with Hybrid Autoregressive Transformer

Official Implementation of the CrossMAE paper: Rethinking Patch Dependence for Masked Autoencoders

[ICLR & NeurIPS 2025] Repository for Show-o series, One Single Transformer to Unify Multimodal Understanding and Generation.