ai-models/Wan2.1-I2V-14B-480P
2
0
Fork 0
mirror of https://modelscope.cn/models/Wan-AI/Wan2.1-I2V-14B-480P synced 2026-04-02 19:12:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
yingda
2025-02-26 01:40:58 +00:00
parent ea4065a014
commit 38eb111376
1 changed files with 268 additions and 259 deletions
Download Patch File Download Diff File Expand all files Collapse all files

527
README.md
View File

@ -1,259 +1,268 @@
# Wan2.1 ---
license: apache-2.0
<p align="center"> language:
<img src="assets/logo.png" width="400"/> - en
<p> - zh
pipeline_tag: image-to-video
<p align="center"> library_name: diffusers
💜 <a href=""><b>Wan</b></a> &nbsp&nbsp ｜ &nbsp&nbsp 🖥️ <a href="https://github.com/Wan-Video/Wan2.1">GitHub</a> &nbsp&nbsp | &nbsp&nbsp🤗 <a href="https://huggingface.co/Wan-AI/">Hugging Face</a>&nbsp&nbsp | &nbsp&nbsp🤖 <a href="https://modelscope.cn/organization/Wan-AI">ModelScope</a>&nbsp&nbsp | &nbsp&nbsp 📑 <a href="">Paper</a> &nbsp&nbsp | &nbsp&nbsp 📑 <a href="">Blog</a> &nbsp&nbsp | &nbsp&nbsp💬 <a href="">WeChat (微信)</a>&nbsp&nbsp | &nbsp&nbsp 📖 <a href="https://discord.gg/p5XbdQV7">Discord</a>&nbsp&nbsp tags:
<br> - video
- video-generation
----- ---
# Wan2.1
[**Wan: Open and Advanced Large-Scale Video Generative Models**]("#") <be>
<p align="center">
In this repository, we present **Wan2.1**, a comprehensive and open suite of video foundation models that pushes the boundaries of video generation. **Wan2.1** offers these key features: <img src="assets/logo.png" width="400"/>
- 👍 **SOTA Performance**: **Wan2.1** consistently outperforms existing open-source models and state-of-the-art commercial solutions across multiple benchmarks. <p>
- 👍 **Supports Consumer-grade GPUs**: The T2V-1.3B model requires only 8.19 GB VRAM, making it compatible with almost all consumer-grade GPUs. It can generate a 5-second 480P video on an RTX 4090 in about 4 minutes (without optimization techniques like quantization). Its performance is even comparable to some closed-source models.
- 👍 **Multiple Tasks**: **Wan2.1** excels in Text-to-Video, Image-to-Video, Video Editing, Text-to-Image, and Video-to-Audio, advancing the field of video generation. <p align="center">
- 👍 **Visual Text Generation**: **Wan2.1** is the first video model capable of generating both Chinese and English text, featuring robust text generation that enhances its practical applications. 💜 <a href=""><b>Wan</b></a> &nbsp&nbsp ｜ &nbsp&nbsp 🖥️ <a href="https://github.com/Wan-Video/Wan2.1">GitHub</a> &nbsp&nbsp | &nbsp&nbsp🤗 <a href="https://huggingface.co/Wan-AI/">Hugging Face</a>&nbsp&nbsp | &nbsp&nbsp🤖 <a href="https://modelscope.cn/organization/Wan-AI">ModelScope</a>&nbsp&nbsp | &nbsp&nbsp 📑 <a href="">Paper (Coming soon)</a> &nbsp&nbsp | &nbsp&nbsp 📑 <a href="https://wanxai.com">Blog</a> &nbsp&nbsp | &nbsp&nbsp💬 <a href="https://gw.alicdn.com/imgextra/i2/O1CN01tqjWFi1ByuyehkTSB_!!6000000000015-0-tps-611-1279.jpg">WeChat Group</a>&nbsp&nbsp | &nbsp&nbsp 📖 <a href="https://discord.gg/p5XbdQV7">Discord</a>&nbsp&nbsp
- 👍 **Powerful Video VAE**: **Wan-VAE** delivers exceptional efficiency and performance, encoding and decoding 1080P videos of any length while preserving temporal information, making it an ideal foundation for video and image generation. <br>
-----
This repo contains our I2V-14B model, which is capable of generating 480P videos, offering advantages in terms of fast generation and excellent quality.
[**Wan: Open and Advanced Large-Scale Video Generative Models**]() <be>
## Video Demos In this repository, we present **Wan2.1**, a comprehensive and open suite of video foundation models that pushes the boundaries of video generation. **Wan2.1** offers these key features:
- 👍 **SOTA Performance**: **Wan2.1** consistently outperforms existing open-source models and state-of-the-art commercial solutions across multiple benchmarks.
<div align="center"> - 👍 **Supports Consumer-grade GPUs**: The T2V-1.3B model requires only 8.19 GB VRAM, making it compatible with almost all consumer-grade GPUs. It can generate a 5-second 480P video on an RTX 4090 in about 4 minutes (without optimization techniques like quantization). Its performance is even comparable to some closed-source models.
<video width="80%" controls> - 👍 **Multiple Tasks**: **Wan2.1** excels in Text-to-Video, Image-to-Video, Video Editing, Text-to-Image, and Video-to-Audio, advancing the field of video generation.
<source src="https://cloud.video.taobao.com/vod/Jth64Y7wNoPcJki_Bo1ZJTDBvNjsgjlVKsNs05Fqfps.mp4" type="video/mp4"> - 👍 **Visual Text Generation**: **Wan2.1** is the first video model capable of generating both Chinese and English text, featuring robust text generation that enhances its practical applications.
Your browser does not support the video tag. - 👍 **Powerful Video VAE**: **Wan-VAE** delivers exceptional efficiency and performance, encoding and decoding 1080P videos of any length while preserving temporal information, making it an ideal foundation for video and image generation.
</video>
</div>
This repo contains our I2V-14B model, which is capable of generating 480P videos, offering advantages in terms of fast generation and excellent quality.
## 🔥 Latest News!!
* Feb 25, 2025: 👋 We've released the inference code and weights of Wan2.1. ## Video Demos
<div align="center">
## 📑 Todo List <video width="80%" controls>
- Wan2.1 Text-to-Video <source src="https://cloud.video.taobao.com/vod/Jth64Y7wNoPcJki_Bo1ZJTDBvNjsgjlVKsNs05Fqfps.mp4" type="video/mp4">
- [x] Multi-GPU Inference code of the 14B and 1.3B models Your browser does not support the video tag.
- [x] Checkpoints of the 14B and 1.3B models </video>
- [x] Gradio demo </div>
- [ ] Diffusers integration
- [ ] ComfyUI integration ## 🔥 Latest News!!
- Wan2.1 Image-to-Video
- [x] Multi-GPU Inference code of the 14B model * Feb 25, 2025: 👋 We've released the inference code and weights of Wan2.1.
- [x] Checkpoints of the 14B model
- [x] Gradio demo
- [ ] Diffusers integration ## 📑 Todo List
- [ ] ComfyUI integration - Wan2.1 Text-to-Video
- [x] Multi-GPU Inference code of the 14B and 1.3B models
- [x] Checkpoints of the 14B and 1.3B models
## Quickstart - [x] Gradio demo
- [ ] Diffusers integration
#### Installation - [ ] ComfyUI integration
Clone the repo: - Wan2.1 Image-to-Video
``` - [x] Multi-GPU Inference code of the 14B model
git clone https://github.com/Wan-Video/Wan2.1.git - [x] Checkpoints of the 14B model
cd Wan2.1 - [x] Gradio demo
``` - [ ] Diffusers integration
- [ ] ComfyUI integration
Install dependencies:
```
# Ensure torch >= 2.4.0 ## Quickstart
pip install -r requirements.txt
``` #### Installation
Clone the repo:
```
#### Model Download git clone https://github.com/Wan-Video/Wan2.1.git
cd Wan2.1
| Models | Download Link | Notes | ```
| --------------|-------------------------------------------------------------------------------|-------------------------------|
| T2V-14B | [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B) | Supports both 480P and 720P Install dependencies:
| I2V-14B-720P | [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-720P) | Supports 720P ```
| I2V-14B-480P | [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-480P) | Supports 480P # Ensure torch >= 2.4.0
| T2V-1.3B | [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B) | Supports 480P pip install -r requirements.txt
```
> 💡Note: The 1.3B model is capable of generating videos at 720P resolution. However, due to limited training at this resolution, the results are generally less stable compared to 480P. For optimal performance, we recommend using 480P resolution.
#### Model Download
Download models using huggingface-cli:
``` | Models | Download Link | Notes |
pip install "huggingface_hub[cli]" | --------------|-------------------------------------------------------------------------------|-------------------------------|
huggingface-cli download --resume-download Wan-AI/Wan2.1-I2V-14B-480P --local-dir ./Wan2.1-I2V-14B-480P | T2V-14B | 🤗 [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B) 🤖 [ModelScope](https://www.modelscope.cn/models/Wan-AI/Wan2.1-T2V-14B) | Supports both 480P and 720P
``` | I2V-14B-720P | 🤗 [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-720P) 🤖 [ModelScope](https://www.modelscope.cn/models/Wan-AI/Wan2.1-I2V-14B-720P) | Supports 720P
| I2V-14B-480P | 🤗 [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-480P) 🤖 [ModelScope](https://www.modelscope.cn/models/Wan-AI/Wan2.1-I2V-14B-480P) | Supports 480P
| T2V-1.3B | 🤗 [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B) 🤖 [ModelScope](https://www.modelscope.cn/models/Wan-AI/Wan2.1-T2V-1.3B) | Supports 480P
#### Run Image-to-Video Generation
> 💡Note: The 1.3B model is capable of generating videos at 720P resolution. However, due to limited training at this resolution, the results are generally less stable compared to 480P. For optimal performance, we recommend using 480P resolution.
Similar to Text-to-Video, Image-to-Video is also divided into processes with and without the prompt extension step. The specific parameters and their corresponding settings are as follows:
<table>
<thead> Download models using huggingface-cli:
<tr> ```
<th rowspan="2">Task</th> pip install "huggingface_hub[cli]"
<th colspan="2">Resolution</th> huggingface-cli download Wan-AI/Wan2.1-I2V-14B-480P --local-dir ./Wan2.1-I2V-14B-480P
<th rowspan="2">Model</th> ```
</tr>
<tr>
<th>480P</th> #### Run Image-to-Video Generation
<th>720P</th>
</tr> Similar to Text-to-Video, Image-to-Video is also divided into processes with and without the prompt extension step. The specific parameters and their corresponding settings are as follows:
</thead> <table>
<tbody> <thead>
<tr> <tr>
<td>i2v-14B</td> <th rowspan="2">Task</th>
<td style="color: green;">❌</td> <th colspan="2">Resolution</th>
<td style="color: green;">✔️</td> <th rowspan="2">Model</th>
<td>Wan2.1-I2V-14B-720P</td> </tr>
</tr> <tr>
<tr> <th>480P</th>
<td>i2v-14B</td> <th>720P</th>
<td style="color: green;">✔️</td> </tr>
<td style="color: red;">❌</td> </thead>
<td>Wan2.1-T2V-14B-480P</td> <tbody>
</tr> <tr>
</tbody> <td>i2v-14B</td>
</table> <td style="color: green;">❌</td>
<td style="color: green;">✔️</td>
<td>Wan2.1-I2V-14B-720P</td>
##### (1) Without Prompt Extention </tr>
<tr>
- Single-GPU inference <td>i2v-14B</td>
``` <td style="color: green;">✔️</td>
python generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside." <td style="color: red;">❌</td>
``` <td>Wan2.1-T2V-14B-480P</td>
</tr>
</tbody>
- Multi-GPU inference using FSDP + xDiT USP </table>
```
pip install "xfuser>=0.4.1" ##### (1) Without Prompt Extention
torchrun --nproc_per_node=8 generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --dit_fsdp --t5_fsdp --ulysses_size 8 --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside."
``` - Single-GPU inference
```
##### (2) Using Prompt Extention python generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside."
```
The process of prompt extension can be referenced [here](#2-using-prompt-extention). > 💡For the Image-to-Video task, the `size` parameter represents the area of the generated video, with the aspect ratio following that of the original input image.
Run with local prompt extention using `Qwen/Qwen2.5-VL-7B-Instruct`: - Multi-GPU inference using FSDP + xDiT USP
```
python generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --use_prompt_extend --prompt_extend_model Qwen/Qwen2.5-VL-7B-Instruct --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside." ```
``` pip install "xfuser>=0.4.1"
torchrun --nproc_per_node=8 generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --dit_fsdp --t5_fsdp --ulysses_size 8 --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside."
Run with remote prompt extention using `dashscope`: ```
```
DASH_API_KEY=your_key python generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --use_prompt_extend --prompt_extend_method 'dashscope' --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside." ##### (2) Using Prompt Extention
```
Run with local prompt extention using `Qwen/Qwen2.5-VL-7B-Instruct`:
##### (3) Runing local gradio ```
python generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --use_prompt_extend --prompt_extend_model Qwen/Qwen2.5-VL-7B-Instruct --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside."
``` ```
cd gradio
# if one only uses 480P model in gradio Run with remote prompt extention using `dashscope`:
DASH_API_KEY=your_key python i2v_14B_singleGPU.py --prompt_extend_method 'dashscope' --ckpt_dir_480p ./Wan2.1-I2V-14B-480P ```
DASH_API_KEY=your_key python generate.py --task i2v-14B --size 832*480 --ckpt_dir ./Wan2.1-I2V-14B-480P --image examples/i2v_input.JPG --use_prompt_extend --prompt_extend_method 'dashscope' --prompt "Summer beach vacation style, a white cat wearing sunglasses sits on a surfboard. The fluffy-furred feline gazes directly at the camera with a relaxed expression. Blurred beach scenery forms the background featuring crystal-clear waters, distant green hills, and a blue sky dotted with white clouds. The cat assumes a naturally relaxed posture, as if savoring the sea breeze and warm sunlight. A close-up shot highlights the feline's intricate details and the refreshing atmosphere of the seaside."
# if one only uses 720P model in gradio ```
DASH_API_KEY=your_key python i2v_14B_singleGPU.py --prompt_extend_method 'dashscope' --ckpt_dir_720p ./Wan2.1-I2V-14B-720P
##### (3) Runing local gradio
# if one uses both 480P and 720P models in gradio
DASH_API_KEY=your_key python i2v_14B_singleGPU.py --prompt_extend_method 'dashscope' --ckpt_dir_480p ./Wan2.1-I2V-14B-480P --ckpt_dir_720p ./Wan2.1-I2V-14B-720P ```
``` cd gradio
# if one only uses 480P model in gradio
DASH_API_KEY=your_key python i2v_14B_singleGPU.py --prompt_extend_method 'dashscope' --ckpt_dir_480p ./Wan2.1-I2V-14B-480P
## Manual Evaluation
# if one only uses 720P model in gradio
We conducted extensive manual evaluations to evaluate the performance of the Image-to-Video model, and the results are presented in the table below. The results clearly indicate that **Wan2.1** outperforms both closed-source and open-source models. DASH_API_KEY=your_key python i2v_14B_singleGPU.py --prompt_extend_method 'dashscope' --ckpt_dir_720p ./Wan2.1-I2V-14B-720P
<div align="center"> # if one uses both 480P and 720P models in gradio
<img src="assets/i2v_res.png" alt="" style="width: 80%;" /> DASH_API_KEY=your_key python i2v_14B_singleGPU.py --prompt_extend_method 'dashscope' --ckpt_dir_480p ./Wan2.1-I2V-14B-480P --ckpt_dir_720p ./Wan2.1-I2V-14B-720P
</div> ```
## Computational Efficiency on Different GPUs ## Manual Evaluation
We test the computational efficiency of different **Wan2.1** models on different GPUs in the following table. The results are presented in the format: **Total time (s) / peak GPU memory (GB)**. We conducted extensive manual evaluations to evaluate the performance of the Image-to-Video model, and the results are presented in the table below. The results clearly indicate that **Wan2.1** outperforms both closed-source and open-source models.
<div align="center">
<div align="center"> <img src="assets/i2v_res.png" alt="" style="width: 80%;" />
<img src="assets/comp_effic.png" alt="" style="width: 80%;" /> </div>
</div>
> The parameter settings for the tests presented in this table are as follows: ## Computational Efficiency on Different GPUs
> (1) For the 1.3B model on 8 GPUs, set `--ring_size 8` and `--ulysses_size 1`;
> (2) For the 14B model on 1 GPU, use `--offload_model True`; We test the computational efficiency of different **Wan2.1** models on different GPUs in the following table. The results are presented in the format: **Total time (s) / peak GPU memory (GB)**.
> (3) For the 1.3B model on a single 4090 GPU, set `--offload_model True --t5_cpu`;
> (4) For all testings, no prompt extension was applied, meaning `--use_prompt_extend` was not enabled.
<div align="center">
------- <img src="assets/comp_effic.png" alt="" style="width: 80%;" />
</div>
## Introduction of Wan2.1
> The parameter settings for the tests presented in this table are as follows:
**Wan2.1** is designed on the mainstream diffusion transformer paradigm, achieving significant advancements in generative capabilities through a series of innovations. These include our novel spatio-temporal variational autoencoder (VAE), scalable training strategies, large-scale data construction, and automated evaluation metrics. Collectively, these contributions enhance the model’s performance and versatility. > (1) For the 1.3B model on 8 GPUs, set `--ring_size 8` and `--ulysses_size 1`;
> (2) For the 14B model on 1 GPU, use `--offload_model True`;
> (3) For the 1.3B model on a single 4090 GPU, set `--offload_model True --t5_cpu`;
##### (1) 3D Variational Autoencoders > (4) For all testings, no prompt extension was applied, meaning `--use_prompt_extend` was not enabled.
We propose a novel 3D causal VAE architecture, termed **Wan-VAE** specifically designed for video generation. By combining multiple strategies, we improve spatio-temporal compression, reduce memory usage, and ensure temporal causality. **Wan-VAE** demonstrates significant advantages in performance efficiency compared to other open-source VAEs. Furthermore, our **Wan-VAE** can encode and decode unlimited-length 1080P videos without losing historical temporal information, making it particularly well-suited for video generation tasks.
-------
<div align="center"> ## Introduction of Wan2.1
<img src="assets/video_vae_res.jpg" alt="" style="width: 80%;" />
</div> **Wan2.1** is designed on the mainstream diffusion transformer paradigm, achieving significant advancements in generative capabilities through a series of innovations. These include our novel spatio-temporal variational autoencoder (VAE), scalable training strategies, large-scale data construction, and automated evaluation metrics. Collectively, these contributions enhance the model’s performance and versatility.
##### (2) Video Diffusion DiT ##### (1) 3D Variational Autoencoders
We propose a novel 3D causal VAE architecture, termed **Wan-VAE** specifically designed for video generation. By combining multiple strategies, we improve spatio-temporal compression, reduce memory usage, and ensure temporal causality. **Wan-VAE** demonstrates significant advantages in performance efficiency compared to other open-source VAEs. Furthermore, our **Wan-VAE** can encode and decode unlimited-length 1080P videos without losing historical temporal information, making it particularly well-suited for video generation tasks.
**Wan2.1** is designed using the Flow Matching framework within the paradigm of mainstream Diffusion Transformers. Our model's architecture uses the T5 Encoder to encode multilingual text input, with cross-attention in each transformer block embedding the text into the model structure. Additionally, we employ an MLP with a Linear layer and a SiLU layer to process the input time embeddings and predict six modulation parameters individually. This MLP is shared across all transformer blocks, with each block learning a distinct set of biases. Our experimental findings reveal a significant performance improvement with this approach at the same parameter scale.
<div align="center"> <div align="center">
<img src="assets/video_dit_arch.jpg" alt="" style="width: 80%;" /> <img src="assets/video_vae_res.jpg" alt="" style="width: 80%;" />
</div> </div>
| Model | Dimension | Input Dimension | Output Dimension | Feedforward Dimension | Frequency Dimension | Number of Heads | Number of Layers | ##### (2) Video Diffusion DiT
|--------|-----------|-----------------|------------------|-----------------------|---------------------|-----------------|------------------|
| 1.3B | 1536 | 16 | 16 | 8960 | 256 | 12 | 30 | **Wan2.1** is designed using the Flow Matching framework within the paradigm of mainstream Diffusion Transformers. Our model's architecture uses the T5 Encoder to encode multilingual text input, with cross-attention in each transformer block embedding the text into the model structure. Additionally, we employ an MLP with a Linear layer and a SiLU layer to process the input time embeddings and predict six modulation parameters individually. This MLP is shared across all transformer blocks, with each block learning a distinct set of biases. Our experimental findings reveal a significant performance improvement with this approach at the same parameter scale.
| 14B | 5120 | 16 | 16 | 13824 | 256 | 40 | 40 |
<div align="center">
<img src="assets/video_dit_arch.jpg" alt="" style="width: 80%;" />
</div>
##### Data
We curated and deduplicated a candidate dataset comprising a vast amount of image and video data. During the data curation process, we designed a four-step data cleaning process, focusing on fundamental dimensions, visual quality and motion quality. Through the robust data processing pipeline, we can easily obtain high-quality, diverse, and large-scale training sets of images and videos. | Model | Dimension | Input Dimension | Output Dimension | Feedforward Dimension | Frequency Dimension | Number of Heads | Number of Layers |
|--------|-----------|-----------------|------------------|-----------------------|---------------------|-----------------|------------------|
![figure1](assets/data_for_diff_stage.jpg "figure1") | 1.3B | 1536 | 16 | 16 | 8960 | 256 | 12 | 30 |
| 14B | 5120 | 16 | 16 | 13824 | 256 | 40 | 40 |
##### Comparisons to SOTA
We compared **Wan2.1** with leading open-source and closed-source models to evaluate the performace. Using our carefully designed set of 1,035 internal prompts, we tested across 14 major dimensions and 26 sub-dimensions. Then we calculated the total score through a weighted average based on the importance of each dimension. The detailed results are shown in the table below. These results demonstrate our model's superior performance compared to both open-source and closed-source models.
##### Data
![figure1](assets/vben_vs_sota.png "figure1")
We curated and deduplicated a candidate dataset comprising a vast amount of image and video data. During the data curation process, we designed a four-step data cleaning process, focusing on fundamental dimensions, visual quality and motion quality. Through the robust data processing pipeline, we can easily obtain high-quality, diverse, and large-scale training sets of images and videos.
## Citation ![figure1](assets/data_for_diff_stage.jpg "figure1")
If you find our work helpful, please cite us.
``` ##### Comparisons to SOTA
@article{wan2.1, We compared **Wan2.1** with leading open-source and closed-source models to evaluate the performace. Using our carefully designed set of 1,035 internal prompts, we tested across 14 major dimensions and 26 sub-dimensions. We then compute the total score by performing a weighted calculation on the scores of each dimension, utilizing weights derived from human preferences in the matching process. The detailed results are shown in the table below. These results demonstrate our model's superior performance compared to both open-source and closed-source models.
title = {Wan: Open and Advanced Large-Scale Video Generative Models},
author = {Wan Team}, ![figure1](assets/vben_vs_sota.png "figure1")
journal = {},
year = {2025}
} ## Citation
``` If you find our work helpful, please cite us.
## License Agreement ```
The models in this repository are licensed under the Apache 2.0 License. We claim no rights over the your generate contents, granting you the freedom to use them while ensuring that your usage complies with the provisions of this license. You are fully accountable for your use of the models, which must not involve sharing any content that violates applicable laws, causes harm to individuals or groups, disseminates personal information intended for harm, spreads misinformation, or targets vulnerable populations. For a complete list of restrictions and details regarding your rights, please refer to the full text of the [license](LICENSE.txt). @article{wan2.1,
title = {Wan: Open and Advanced Large-Scale Video Generative Models},
author = {Wan Team},
## Acknowledgements journal = {},
year = {2025}
We would like to thank the contributors to the [SD3](https://huggingface.co/stabilityai/stable-diffusion-3-medium), [QWen](https://huggingface.co/Qwen), [umt5-xxl](https://huggingface.co/google/umt5-xxl), [diffusers](https://github.com/huggingface/diffusers) and [HuggingFace](https://huggingface.co) repositories, for their open research and exploration. }
```
## License Agreement
## Contact Us The models in this repository are licensed under the Apache 2.0 License. We claim no rights over the your generate contents, granting you the freedom to use them while ensuring that your usage complies with the provisions of this license. You are fully accountable for your use of the models, which must not involve sharing any content that violates applicable laws, causes harm to individuals or groups, disseminates personal information intended for harm, spreads misinformation, or targets vulnerable populations. For a complete list of restrictions and details regarding your rights, please refer to the full text of the [license](LICENSE.txt).
If you would like to leave a message to our research or product teams, feel free to join our [Discord](https://discord.gg/p5XbdQV7) or [WeChat groups]()!
## Acknowledgements
We would like to thank the contributors to the [SD3](https://huggingface.co/stabilityai/stable-diffusion-3-medium), [Qwen](https://huggingface.co/Qwen), [umt5-xxl](https://huggingface.co/google/umt5-xxl), [diffusers](https://github.com/huggingface/diffusers) and [HuggingFace](https://huggingface.co) repositories, for their open research.
## Contact Us
If you would like to leave a message to our research or product teams, feel free to join our [Discord](https://discord.gg/p5XbdQV7) or [WeChat groups](https://gw.alicdn.com/imgextra/i2/O1CN01tqjWFi1ByuyehkTSB_!!6000000000015-0-tps-611-1279.jpg)!