Official code for paper Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models

Scalable group inference for generating high quality and diverse images with diffusion models.

LoRA fine-tuning for FLUX.2 to improve virtual try-on (VTON) capabilities

Implementation of Particle Guidance: non-I.I.D. Diverse Sampling with Diffusion Models

[Tutorial] Few-Step Distillation for Text-to-Image Generation: A Practical Guide

Cosmos-Predict2.5, the latest version of the Cosmos World Foundation Models (WFMs) family, specialized for simulating and predicting the future state of the world in the form of video.

Official implementation for What matters for Representation Alignment: Global Information or Spatial Structure?

Implementation of "Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length"

Official PyTorch Implementation of "Optimal Stepsize for Diffusion Sampling".

PyTorch implementation of JiT https://arxiv.org/abs/2511.13720

Qwen2.5-Omni is an end-to-end multimodal model by Qwen team at Alibaba Cloud, capable of understanding text, audio, vision, video, and performing real-time speech generation.

Enhancing Reward Models for High-quality Image Generation: Beyond Text-Image Alignment [ICCV 2025] - Official implementation

Official PyTorch Implementation of "Diffusion Transformers with Representation Autoencoders"

SigLIP-based Aesthetic Score Predictor

Qwen-Image is a powerful image generation foundation model capable of complex text rendering and precise image editing.

DiffusionNFT: Online Diffusion Reinforcement with Forward Process

Control and limit battery charging on Apple Silicon MacBooks.

Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and video, as well as generating speech in real time.

Directly Aligning the Full Diffusion Trajectory with Fine-Grained Human Preference

An official implementation of Coefficients-Preserving Sampling for Reinforcement Learning with Flow Matching

A unified inference and post-training framework for accelerated video generation.

The official UniVerse-1 code.

Official implementation of Pref-GRPO: Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning

NextStep-1: SOTA Autogressive Image Generation with Continuous Tokens. A research project developed by the StepFun’s Multimodal Intelligence team.

Pytorch implementation for MeanFlow

TempFlow-GRPO (Temporal Flow GRPO), a principled GRPO framework that captures and exploits the temporal structure inherent in flow-based generation.

Enjoy the magic of Diffusion models!