Real-time interactive streaming digital human enables synchronous audio and video dialogue. It can basically achieve commercial effects.

Effect of wav2lip | Effect of ernerf | Effect of musetalk

• December 8, 2024: Improved multi-concurrency, and the video memory does not increase with the number of concurrent connections.
• December 21, 2024: Added model warm-up for wav2lip and musetalk to solve the problem of stuttering during the first inference. Thanks to @heimaojinzhangyz
• December 28, 2024: Added the digital human model Ultralight-Digital-Human. Thanks to @lijihua2017
• February 7, 2025: Added fish-speech tts
• February 21, 2025: Added the open-source model wav2lip256. Thanks to @不蠢不蠢
• March 2, 2025: Added Tencent's speech synthesis service
• March 16, 2025: Supports mac gpu inference. Thanks to @GcsSloop

1. Supports multiple digital human models: ernerf, musetalk, wav2lip, Ultralight-Digital-Human
2. Supports voice cloning
3. Supports interrupting the digital human while it is speaking
4. Supports full-body video stitching
5. Supports rtmp and webrtc
6. Supports video arrangement: Play custom videos when not speaking
7. Supports multi-concurrency

Tested on Ubuntu 20.04, Python 3.10, Pytorch 1.12 and CUDA 11.3

conda create -n nerfstream python=3.10
conda activate nerfstream
# If the cuda version is not 11.3 (confirm the version by running nvidia-smi), install the corresponding version of pytorch according to <https://pytorch.org/get-started/previous-versions/> 
conda install pytorch==1.12.1 torchvision==0.13.1 cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt
# If you need to train the ernerf model, install the following libraries
# pip install "git+https://github.com/facebookresearch/pytorch3d.git"
# pip install tensorflow-gpu==2.8.0
# pip install --upgrade "protobuf<=3.20.1"

Common installation issues FAQ
For setting up the linux cuda environment, you can refer to this article https://zhuanlan.zhihu.com/p/674972886

• Download the models
  Quark Cloud Disk https://pan.quark.cn/s/83a750323ef0
  Google Drive https://drive.google.com/drive/folders/1FOC_MD6wdogyyX_7V1d4NDIO7P9NlSAJ?usp=sharing
  Copy wav2lip256.pth to the models folder of this project and rename it to wav2lip.pth;
  Extract wav2lip256_avatar1.tar.gz and copy the entire folder to the data/avatars folder of this project.

• Run
  python app.py --transport webrtc --model wav2lip --avatar_id wav2lip256_avatar1
  Open http://serverip:8010/webrtcapi.html in a browser. First click'start' to play the digital human video; then enter any text in the text box and submit it. The digital human will broadcast this text.
  The server side needs to open ports tcp:8010; udp:1-65536
  If you need to purchase a high-definition wav2lip model for commercial use, Link.

• Quick experience
  https://www.compshare.cn/images-detail?ImageID=compshareImage-18tpjhhxoq3j&referral_code=3XW3852OBmnD089hMMrtuU&ytag=GPU_GitHub_livetalking1.3 Create an instance with this image to run it.

If you can't access huggingface, before running

export HF_ENDPOINT=https://hf-mirror.com

Usage instructions: https://livetalking-doc.readthedocs.io/en/latest

No need for the previous installation, just run directly.

docker run --gpus all -it --network=host --rm registry.cn-beijing.aliyuncs.com/codewithgpu2/lipku-metahuman-stream:2K9qaMBu8v

The code is in /root/metahuman-stream. First, git pull to get the latest code, and then execute the commands as in steps 2 and 3.

The following images are provided:

• autodl image: https://www.codewithgpu.com/i/lipku/metahuman-stream/base
  autodl Tutorial
• ucloud image: https://www.compshare.cn/images-detail?ImageID=compshareImage-18tpjhhxoq3j&referral_code=3XW3852OBmnD089hMMrtuU&ytag=GPU_livetalking1.3
  Any port can be opened, and there is no need to deploy an srs service additionally.
  ucloud Tutorial

• Added chatgpt to enable digital human dialogue
• Voice cloning
• Replace the digital human with a video when it is silent
• MuseTalk
• Wav2Lip
• Ultralight-Digital-Human

---

If this project is helpful to you, please give it a star. Friends who are interested are also welcome to join in and improve this project together.

• Knowledge Planet: https://t.zsxq.com/7NMyO, where high-quality common problems, best practice experiences, and problem solutions are accumulated.
• WeChat Official Account: Digital Human Technology