Cosmos-Transfer2.5, built on top of Cosmos-Predict2.5, produces high-quality world simulations conditioned on multiple spatial control inputs.

🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.

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.

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

Cosmos-Reason1 models understand the physical common sense and generate appropriate embodied decisions in natural language through long chain-of-thought reasoning processes.

A generative world for general-purpose robotics & embodied AI learning.

Making large AI models cheaper, faster and more accessible

Native Multimodal Models are World Learners

📹 A more flexible framework that can generate videos at any resolution and creates videos from images.

verl: Volcano Engine Reinforcement Learning for LLMs

A Paper List for Humanoid Robot Learning.

[ICCV 2025 Highlights] Large-scale photo-realistic virtual worlds for embodied AI

Official repo for paper "Structured 3D Latents for Scalable and Versatile 3D Generation" (CVPR'25 Spotlight).

rCM: SOTA Diffusion Distillation & Few-Step Video Generation

[NeurIPS 2025] An official implementation of Flow-GRPO: Training Flow Matching Models via Online RL

text and image to video generation: CogVideoX (2024) and CogVideo (ICLR 2023)

LongLive: Real-time Interactive Long Video Generation

[NeurIPS 2025 D&B] Open-source Multi-agent Poster Generation from Papers

A growing curation of Text-to-3D, Diffusion-to-3D works.

GLOMAP - Global Structured-from-Motion Revisited

Cosmos-Transfer1 is a world-to-world transfer model designed to bridge the perceptual divide between simulated and real-world environments.

[ICLR 2025] Autoregressive Video Generation without Vector Quantization

Cosmos-Predict2 is a collection of general-purpose world foundation models for Physical AI that can be fine-tuned into customized world models for downstream applications.

High-Resolution 3D Assets Generation with Large Scale Hunyuan3D Diffusion Models.

Awesome-LLM-3D: a curated list of Multi-modal Large Language Model in 3D world Resources

[IROS 2025 Award Finalist] The Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems

Open-source unified multimodal model

[ICLR 2024 Spotlight] SyncDreamer: Generating Multiview-consistent Images from a Single-view Image

[NeurIPS 2025] WorldMem: Long-term Consistent World Simulation with Memory

[CVPR 2024] RoMa: Robust Dense Feature Matching; RoMa is the robust dense feature matcher capable of estimating pixel-dense warps and reliable certainties for almost any image pair.