ComfyUI_EchoMimic

You can use EchoMimic & EchoMimic V2 & EchoMimic V3 in comfyui.
Echomimic：Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning
Echomimic_v2: Towards Striking, Simplified, and Semi-Body Human Animation
Echomimic_v3：1.3B Parameters are All You Need for Unified Multi-Modal and Multi-Task Human Animation

---

New Updates

• Support v3 flash model,now you can infer 768x768 size in 12G Vram use 8 steps，768x512 need 6.5G (need new transformer and audio model)，need ‘ block_offload ’ on;
• 支持v3 flash模型，并加以优化，12G，8步可以推理768x768的视频，6.5G可以推理768X512的视频,需要开启block_offload块卸载功能;

Previous

• you can set lowram 'False ' to close mmgp 's fp8 quantum ，will get more quality output./设置lowram为false时，关闭mmgp的FP8 量化以得到更好的质量。
• add LCM support ,if set step=4（and lightX2V lora）,will run in LCM/ 步数设置为4时，自动开启LCM，当然也要lora
• v3版本新增lightX2V Lora的支持， step可以设置为10步（使用Lora时自动开启Unip）/you can use lightX2V Lora when use V3 version, set step=10;
• 修复bug，retina-face 模型改成本地运行
• V3正式上线，测试环境12G VRAM，OOM需要减少视频分块(partial_video_length)的数值，12G可以跑65，16可以试试97，更高可以试试113
• V3 is Done,you can try it now.. need 8G and more (use mmgp,LOW LOW,partial_video_length==65 or 33)

1. Installation

In the ./ComfyUI /custom_nodes directory, run the following:

git clone https://github.com/smthemex/ComfyUI_EchoMimic.git

---

2. Requirements

pip install -r requirements.txt

Notice

---

If use v1 version 如果要使用V1版本：

pip install --no-deps facenet-pytorch 

If use v3 version 如果要使用V3版本： # v3 flash do not use it #flash模型不用

pip install retina-face==0.0.17 #使用须外网下载模型，待处理
pip install mmgp # optional 可选 
pip install tensorflow==2.15.0   #高版本可能会报错，存疑   

• 如果ffmpeg 报错，if ffmpeg error：

pip uninstall ffmpeg   
pip install ffmpeg-python  

• 其他库缺啥装啥。。。
• If the module is missing, , pip install missing module.

3. Models Required

---

3.1 v3 version
3.1.1 from Wan2.1-Fun-V1.1-1.3B-InPdownlaod Wan2.1_VAE.pth and diffusion_pytorch_model.safetensors v3 and v3 flash 3.1.2 use comfyui ,clipvison-h and umt5_xxl_fp8_e4m3fn_scaled.safetensors v3 and v3 flash 3.1.3 wav2vec2-base-960h v3 only 3.1.4 BadToBest/EchoMimicV3 transformer v3 only
3.1.5 retinaface.h5 目录下没有一般会自动下载 v3 only 3.1.6 可选/optional lora kijai v3 only 3.1.7 BadToBest/EchoMimicV3/echomimicv3-flash-pro v3 flash 3.1.8 chinese-wav2vec2-base v3 flash

├── ComfyUI/models/echo_mimic/transformer 
|         ├── diffusion_pytorch_model.safetensors  # Wan2.1-Fun-V1.1-1.3B-InP transformer #3.13G 务必注意模型同名。  v3 and v3 flash
|         ├── config.json
├── ComfyUI/models/echo_mimic/wav2vec2-base-960h   # v3 only 
|         ├── all config json files 
|         ├──  model.safetensors
├── ComfyUI/models/clip
|         ├── umt5_xxl_fp8_e4m3fn_scaled.safetensors   # v3 and v3 flash
├── ComfyUI/models/clip_vision  # v3 and v3 flash 
|         ├──clipvison-h # 1.26G
├── ComfyUI/models/echo_mimic/
|         ├──diffusion_pytorch_model.safetensors  # BadToBest/EchoMimicV3   v3 only 
├── ComfyUI/models/echo_mimic/echomimicv3-flash-pro/
|         ├──diffusion_pytorch_model.safetensors  # BadToBest/EchoMimicV3  v3 flashonly 
├── ComfyUI/models/echo_mimic/chinese-wav2vec2-base/    #v3 flashonly 
|         ├──chinese-wav2vec2-base-fairseq-ckpt.pt 
|         ├──model.safetensors
|         ├──all config      
├── ComfyUI/models/vae
|         ├── Wan2.1_VAE.pth    # v3 and v3 flash 
├── ComfyUI/models/echo_mimic/.deepface/weights/    #注意.deepface前面有个点，这个是方便不能翻墙玩家  #  v3 only 
|         ├──retinaface.h5
├── ComfyUI/models/loras/    
|         ├──lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors  #KJ #  v3 only 

3.2 V1 & V2 Shared model v1 和 v2 共用的模型:
如果能直连抱脸,点击就会自动下载所需模型,不需要手动下载.

• unet link
• V1 & V2 audio link
• vae(stabilityai/sd-vae-ft-mse) link

├── ComfyUI/models/ echo_mimic
|         ├── unet
|             ├── diffusion_pytorch_model.bin
|             ├── config.json
|         ├── audio_processor
|             ├── whisper_tiny.pt
├── ComfyUI/models/vae
|             ├── diffusion_pytorch_model.safetensors or rename sd-vae-ft-mse.safetensors

3.3 V1 models V1使用以下模型:

• V1 address link
• Audio-Drived Algo Inference 音频驱动

├── ComfyUI/models/echo_mimic
|         ├── denoising_unet.pth
|         ├── face_locator.pth
|         ├── motion_module.pth
|         ├── reference_unet.pth
Audio-Drived Algo Inference  acc  音频驱动加速版
|         ├── denoising_unet_acc.pth
|         ├── motion_module_acc.pth

• Using Pose-Drived Algo Inference 姿态驱动

├── ComfyUI/models/echo_mimic
|         ├── denoising_unet_pose.pth
|         ├── face_locator_pose.pth
|         ├── motion_module_pose.pth
|         ├── reference_unet_pose.pth
Using Pose-Drived Algo Inference  ACC   姿态驱动加速版
|         ├── denoising_unet_pose_acc.pth
|         ├── motion_module_pose_acc.pth

3.4 v2 version
use model below V2, Automatic download, you can manually add it 使用以下模型,使用及自动下载,你可以手动添加:
模型地址address:huggingface

├── ComfyUI/models/echo_mimic/v2
|         ├── denoising_unet.pth
|         ├── motion_module.pth
|         ├── pose_encoder.pth
|         ├── reference_unet.pth
if use acc 姿态驱动加速版   
|         ├── denoising_unet_acc.pth
|         ├── motion_module_acc.pth

YOLOm8 download link
sapiens pose download link
sapiens的pose 模型可以量化为fp16的，详细见我的sapiens插件 地址

├── ComfyUI/models/echo_mimic
|         ├── yolov8m.pt
|         ├── sapiens_1b_goliath_best_goliath_AP_639_torchscript.pt2  or/或者 sapiens_1b_goliath_best_goliath_AP_639_torchscript_fp16.pt2

4 Example

---

• V3 flash version

  

• V3 version

  

• V2 version

• V2加载自定义视频驱动视频，V2 loads custom video driver videos

  

• Echomimic_v2 use default pose new version 使用官方默认的pose文件

  

• V1 version

• audio driver 音频驱动

  

• 示例的 VH node : ComfyUI-VideoHelperSuite

---

5 Function Description

---

• v1 版本的生成模式
  a. 单纯音频驱动视频生成模式，infer_mode可选常规的“audio_drived” 和加速版"audio_drived_acc" 模型；
  b. pose驱动生成模式，常规选项为pose_normal_sapiens/pose_normal_dwpose（等同） 加速版本为"pose_acc"模型；
  ----motion_sync：pose驱动时，如果打开且video_file有视频文件时，生成pkl文件，并生成参考视频的视频；pkl文件在input\tensorrt_lite 目录下，再次使用需要重启comfyUI。
  ----motion_sync：如果关闭且pose_dir不为none的时候，读取选定的pose_dir目录名的pkl文件，生成pose视频；如果pose_dir为空的时候，生成基于默认assets\test_pose_demo_pose的视频

• v2 版本的生成模式
  a. infer_mode选择audio_drive,pose_dir 选择列表里的几个默认pose，则使用默认的npy pose文件;
  b. infer_mode选择audio_drive,pose_dir 选择已有的npy文件夹（位于...ComfyUI/input/tensorrt_lite目录下);
  c. infer_mode选择pose_normal_dwpose 或pose_normal_sapiens,video_images连接视频入口，确认...ComfyUI/models/echo_mimic 下有yolov8m.pt 和sapiens_1b_goliath_best_goliath_AP_639_torchscript.pt2 模型,根据输入视频生成npy文件（可以下次用）和视频

• v3 版本生成模式
  a. 基于retina-face库生成
  b. 如果retina-face调用失败，则以默认的女性face作为mask

特别的选项：

• save_video：如果不想使用VH节点时，可以开启，默认关闭；
• draw_mouse：你可以试试；
• length：帧数，时长等于length/fps；
• acc模型 ，6步就可以，但是质量略有下降；
• lowvram :低显存用户可以开启 lowvram users can enable it
• 内置内置图片等比例裁切。
• facecrop_ratio的放大系数为1/facecrop_ratio，如果设置为0.5，面部会得到2倍的放大，建议只在参考图片或者驱动视频中的人脸占比很小的时候，才将facecrop_ratio调整为较小的值.为1 或者0 时不裁切
• cfg数值设置为1，仅在turbo模式有效，其他会报错。V2推荐2.5 V3推荐3.5
• use_mmgp 仅V3版本有效
• partial_video_length 仅V3版本有效，数值越低显存占用越低；
• teacache 仅V3版本有效；

---

6 Citation

EchoMimici

@misc{chen2024echomimic,
  title={EchoMimic: Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning},
  author={Zhiyuan Chen, Jiajiong Cao, Zhiquan Chen, Yuming Li, Chenguang Ma},
  year={2024},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}

EchoMimici-V2

@misc{meng2024echomimic,
  title={EchoMimicV2: Towards Striking, Simplified, and Semi-Body Human Animation},
  author={Rang Meng, Xingyu Zhang, Yuming Li, Chenguang Ma},
  year={2024},
  eprint={2411.10061},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}

EchoMimici-V3

@misc{meng2025echomimicv3,
  title={EchoMimicV3: 1.3B Parameters are All You Need for Unified Multi-Modal and Multi-Task Human Animation},
  author={Rang Meng, Yan Wang, Weipeng Wu, Ruobing Zheng, Yuming Li, Chenguang Ma},
  year={2025},
  eprint={2507.03905},
  archivePrefix={arXiv}
}

LightX2V

@misc{lightx2v,
 author = {LightX2V Contributors},
 title = {LightX2V: Light Video Generation Inference Framework},
 year = {2025},
 publisher = {GitHub},
 journal = {GitHub repository},
 howpublished = {\url{https://github.com/ModelTC/lightx2v}},
}

sapiens

@article{khirodkar2024sapiens,
  title={Sapiens: Foundation for Human Vision Models},
  author={Khirodkar, Rawal and Bagautdinov, Timur and Martinez, Julieta and Zhaoen, Su and James, Austin and Selednik, Peter and Anderson, Stuart and Saito, Shunsuke},
  journal={arXiv preprint arXiv:2408.12569},
  year={2024}
}