Update README.md

README.md

@@ -12,10 +12,10 @@ library_name: diffusers
   <a href="https://github.com/SkyworkAI/Matrix-Game">
     <img src="https://img.shields.io/badge/GitHub-100000?style=flat&logo=github&logoColor=white" alt="GitHub">
   </a>
-  <a href="https://arxiv.org/pdf/2508.13009">
+  <a href="https://github.com/SkyworkAI/Matrix-Game/blob/main/Matrix-Game-3/assets/pdf/report.pdf">
     <img src="https://img.shields.io/badge/arXiv-Report-b31b1b?style=flat&logo=arxiv&logoColor=white" alt="report">
   </a>
-    <a href="https://matrix-game-v2.github.io/">
+  <a href="https://matrix-game-v2.github.io/">
     <img src="https://img.shields.io/badge/Project%20Page-grey?style=flat&logo=huggingface&color=FFA500" alt="Project Page">
   </a>
 
@@ -29,4 +29,56 @@ library_name: diffusers
 - 🚀 **Feature 1**: **Upgraded Data Engine**: Combines Unreal Engine-based synthetic data, large-scale automated AAA game data, and real-world video augmentation to generate high-quality Video–Pose–Action–Prompt data. 
 - 🖱️ **Feature 2**: **Long-horizon Memory & Consistency**: Uses prediction residuals and frame re-injection for self-correction, while camera-aware memory ensures long-term spatiotemporal consistency. 
 - 🎬 **Feature 3**: **Real-Time Interactivity & Open Access**: It employs a multi-segment autoregressive distillation strategy based on Distribution Matching Distillation (DMD), combined with model quantization and VAE decoder distillation to support [40fps] real-time generation at 720p resolution with a 5B model, while maintaining stable memory consistency over minute-long sequence.
-- 👍 **Feature 3**: **Scale Up 28B-MoE Model**: Scaling up to a 2×14B model further improves generation quality, dynamics, and generalization. 
+- 👍 **Feature 3**: **Scale Up 28B-MoE Model**: Scaling up to a 2×14B model further improves generation quality, dynamics, and generalization.
+
+## 🔥 Latest Updates
+
+* [2026-03] 🎉 Initial release of Matrix-Game-3.0 Model
+
+## 🚀 Quick Start
+### Installation
+Create a conda environment and install dependencies:
+```
+conda create -n matrix-game-3.0 python=3.12 -y
+conda activate matrix-game-3.0
+# install FlashAttention
+# Our project also depends on [FlashAttention](https://github.com/Dao-AILab/flash-attention)
+git clone https://github.com/SkyworkAI/Matrix-Game-3.0.git
+cd Matrix-Game-3.0
+pip install -r requirements.txt
+```
+
+### Model Download
+```
+pip install "huggingface_hub[cli]"
+huggingface-cli download Matrix-Game-3.0 --local-dir Matrix-Game-3.0
+```
+### Inference
+Before running inference, you need to prepare:
+- Input image
+- Text prompt
+
+After downloading pretrained models, you can use the following command to generate an interactive video with random actions:
+``` sh
+torchrun --nproc_per_node=$NUM_GPUS generate.py --size 704*1280 --dit_fsdp --t5_fsdp --ckpt_dir Matrix-Game-3.0 --fa_version 3 --use_int8 --num_iterations 12 --num_inference_steps 3 --image demo_images/000/image.png --prompt "a vintage gas station with a classic car parked under a canopy, set against a desert landscape." --save_name test --seed 42 --compile_vae --lightvae_pruning_rate 0.5 --vae_type mg_lightvae --output_dir ./output
+# "num_iterations" refers to the number of iterations you want to generate. The total number of frames generated is given by:57 + (num_iterations - 1) * 40 
+```
+Tips: 
+If you want to use the base model, you can use "--use_base_model --num_inference_steps 50". Otherwise if you want to generating the interactive videos with your own input actions, you can use "--interactive".
+With multiple GPUs, you can pass `--use_async_vae --async_vae_warmup_iters 1` to speed up inference (see [`test.sh`](test.sh)).
+
+## ⭐ Acknowledgements
+- [Diffusers](https://github.com/huggingface/diffusers) for their excellent diffusion model framework
+- [Self-Forcing](https://github.com/guandeh17/Self-Forcing) for their excellent work
+- [GameFactory](https://github.com/KwaiVGI/GameFactory) for their idea of action control module
+- [LightX2V](https://github.com/ModelTC/lightx2v) for their excellent quantization framework
+- [Wan2.2](https://github.com/Wan-Video/Wan2.2) for their strong base model
+- [lingbot-world](https://github.com/Robbyant/lingbot-world) for their context parallel framework 
+## 📜 License
+This project is licensed under the Apache License, Version 2.0 — see [LICENSE.txt](LICENSE.txt).
+
+## 📖 Citation
+If you find this work useful for your research, please kindly cite our paper:
+
+```
+```