--- title: HelloMeme canonical_url: "https://www.modelscope.cn/studios/songkey/HelloMeme" md_url: "https://www.modelscope.cn/studios/songkey/HelloMeme.md" repository: songkey/HelloMeme last_updated: 2025-08-20 sdk_type: gradio sdk_version: 5.12.0 downloads: 0 stars: 70 --- # HelloMeme > HelloMeme - songkey 在 ModelScope 创建的在线 Demo。HelloMeme: Integrating Spatial Knitting Attentions to Embed High-Level and Fidelity-Rich Conditions in Diffusion Models songkey/HelloMeme 是 ModelScope 魔搭社区上的在线可交互 Demo（创空间），基于 gradio 5.12.0 构建。 - **Repository**: songkey/HelloMeme - **SDK**: gradio - **SDK version**: 5.12.0 - **Downloads**: 0 - **Stars**: 70 - **Last updated**: 2025-08-20 Source: https://www.modelscope.cn/studios/songkey/HelloMeme ---

HelloMeme: Integrating Spatial Knitting Attentions to Embed High-Level and Fidelity-Rich Conditions in Diffusion Models

Shengkai Zhang, Nianhong Jiao, Tian Li, Chaojie Yang, Chenhui Xue^*, Boya Niu^*, Jun Gao

HelloVision | HelloGroup Inc.

^* Intern

showcase

## 🔆 New Features/Updates - ☐ [`ExperimentsOnSKAttentions`](https://github.com/HelloVision/ExperimentsOnSKAttentions) for ablation experiments. - ✅ `12/13/2024` Rewrite the code for the Gradio app. - ✅ `12/12/2024` Added HelloMeme V2 (synchronize code from the [`ComfyUI`](https://github.com/HelloVision/ComfyUI_HelloMeme) repo). - ✅ `11/14/2024` Added the `HMControlNet2` module - ✅ `11/12/2024` Added a newly fine-tuned version of [`Animatediff`](https://huggingface.co/songkey/hm_animatediff_frame12) with a patch size of 12, which uses less VRAM (Tested on 2080Ti). - ✅ `11/5/2024` [`ComfyUI`](https://github.com/HelloVision/ComfyUI_HelloMeme) interface for HelloMeme. - ✅ `11/1/2024` Release the code for the core functionalities.. ## Introduction This repository contains the official code implementation of the paper [`HelloMeme`](https://arxiv.org/pdf/2410.22901). Any updates related to the code or models from the paper will be posted here. The code for the ablation experiments discussed in the paper will be added to the [`ExperimentsOnSKAttentions`](https://github.com/HelloVision/ExperimentsOnSKAttentions) section. Additionally, we plan to release a `ComfyUI` interface for HelloMeme, with updates posted here as well. ## Getting Started ### 1. Create a Conda Environment ```bash conda create -n hellomeme python=3.10.11 conda activate hellomeme ``` ### 2. Install PyTorch and FFmpeg To install the latest version of PyTorch, please refer to the official [PyTorch](https://pytorch.org/get-started/locally/) website for detailed installation instructions. Additionally, the code will invoke the system's ffmpeg command for video and audio editing, so the runtime environment must have ffmpeg pre-installed. For installation guidance, please refer to the official [FFmpeg](https://ffmpeg.org/) website. ### 3. Install dependencies ```bash pip install diffusers transformers einops scipy opencv-python tqdm pillow onnxruntime-gpu onnx safetensors accelerate peft imageio imageio[ffmpeg] torchvision ``` > [!IMPORTANT] > > Note the version of diffusers required: frequent updates to diffusers may lead to dependency conflicts. We will periodically check the repo’s compatibility with the latest diffusers version. The currently tested and supported version is **diffusers==0.31.0**. ### 4. Clone the repository ```bash git clone https://github.com/HelloVision/HelloMeme cd HelloMeme ``` ### 5. Run the code ```bash python inference_image.py # for image generation python inference_video.py # for video generation ``` ### 6. Install for Gradio App We recommend setting up the environment with conda. ```bash pip install gradio python app.py ``` After run the app, all models will be downloaded. ## Examples ### Image Generation The input for the image generation script `inference_image.py` consists of a reference image and a drive image, as shown in the figure below:

Reference Image

Drive Image

The output of the image generation script is shown below:

Based on SD1.5

Based on disneyPixarCartoon

### Video Generation The input for the video generation script `inference_video.py` consists of a reference image and a drive video, as shown in the figure below:

Reference Image

Drive Video

The output of the video generation script is shown below:

Based on epicrealism

Based on disneyPixarCartoon

> [!Note] > > If the face in the driving video has significant movement (such as evident camera motion), it is recommended to set the `trans_ratio` parameter to 0 to prevent distorted outputs. > >`inference_video(engines, ref_img_path, drive_video_path, save_path, trans_ratio=0.0)` ## Pretrained Models Our models are all hosted on [🤗](https://huggingface.co/songkey), and the startup script will download them automatically. The specific model information is as follows: | model | size | url | Info | |-------|-------|------|-------------------------------------------------------| | songkey/hm_reference | 312M |

| Face Landmarks (222 points) | Our pipeline also supports loading stylized base models (safetensors). For video generation tasks, using some customized models for portrait generation, such as [**Realistic Vision V6.0 B1**](https://civitai.com/models/4201/realistic-vision-v60-b1), can produce better results. You can download checkpoints and loras into the directories `pretrained_models/` and `pretrained_models/loras/`, respectively. ## Acknowledgements Thanks to 🤗 for providing [diffusers](https://huggingface.co/docs/diffusers), which has greatly enhanced development efficiency in diffusion-related work. We also drew considerable inspiration from [MagicAnimate](https://github.com/magic-research/magic-animate) and [EMO](https://github.com/HumanAIGC/EMO), and [Animatediff](https://github.com/guoyww/AnimateDiff) allowed us to implement the video version at a very low cost. Finally, we thank our colleagues **Shengjie Wu** and **Zemin An**, whose foundational modules played a significant role in this work. ## Citation ```bibtex @misc{zhang2024hellomemeintegratingspatialknitting, title={HelloMeme: Integrating Spatial Knitting Attentions to Embed High-Level and Fidelity-Rich Conditions in Diffusion Models}, author={Shengkai Zhang and Nianhong Jiao and Tian Li and Chaojie Yang and Chenhui Xue and Boya Niu and Jun Gao}, year={2024}, eprint={2410.22901}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2410.22901}, } ``` ## Contact **Shengkai Zhang** (songkey@pku.edu.cn)