[NeurIPS 2025] Flow x RL. "ReinFlow: Fine-tuning Flow Policy with Online Reinforcement Learning". Support VLAs e.g., Pi0, Pi0.5, GR00TN1.5. Fully open-sourced.

(ICASSP 2025, official code)FlowSE: Flow Matching-based Speech Enhancement

Official repository for FlowSE (Interspeech 2025)

unofficial Split Mean Flow Implementation from bytedance

An upscaler node for flow-matching models like Qwen, applying the DemoFusion approach

Diffusion Model Experiment: Fitting an Elliptical Distribution, Flow Matching Proves More Efficient than DDPM The experiment compares training efficiency and sampling trajectories between DDPM (trained with DDPM and sampled with DDIM) and Flow Matching. It highlights the differences in performance during the model training process.

Code for the paper "3D FlowMatch Actor: Unified 3D Policy for Single- and Dual-Arm Manipulation"

A simple toy training code to help understand the difference between diffusion and flow matching

A repo where I play with conditional flow approaches for learning time-varying vector-fields.

Diffusion-SDPO: Safeguarded Direct Preference Optimization for Diffusion Models

No description available

Official repository of FlowInOne: Unifying Multimodal Generation as Image-In Image-Out Flow Matching

Testing flow matching in Euclidean space and Lie groups.

Conditional Flow matching for Image and video

No description available

Implementation of flow matching on tabular data using XGBoost

🎨 Enhance image quality with the FlowMatch Euler Discrete Scheduler, offering sharper, noise-free results in ComfyUI's KSampler.

A toy implementation of a diffusion transformer based "world model" trained on the game of pong or doom.

No description available

A ComfyUI custom node that allows you to invert flow matching model sampling to rewind the clock back to noise for interesting effects.

Text prompt-based voice quality designer using FlowMatching + GPT-SoVITS

Unsupervised Anomaly Detection for Autonomous Driving using Manifold-Aware Conditional Flow Matching on the Waymo Open Motion Dataset.

[IROS 2025] PainDiffusion: Learning to Express Pain - IROS Best Entertainment and Amusement Paper Award

Python library designed to make model experimentation seamless and fast. The goal was simple: treat every component (attention heads, MLPs, MoE layers) as a plug-and-play block so you can focus on the architecture, not the boilerplate.

No description available

A basic implementation of Flow Matching for unconditional 3D point clouds generation.

Official Code for ICLR'26 Work LapFlow

flowmatching+DiT

Flow matching implementation for MNIST digit generation

No description available