Avatars for Skype and Zoom. Democratized.
This app is tested only on GPU and Linux (Ubuntu 18.04) and Mac (low FPS).
It is supposed that there is only one web cam connected to the computer and it's in /dev/video0
. Installation process will create a virtual camera /dev/video1
, so this device must be reserved.
source scripts/install.sh
bash scripts/download_data.sh
For Mac it's quite difficult to create a virtual camera, so we'll use CamTwist app:
- Download and install CamTwist from here. It's easy.
- Setup
avatarify
conda environment with all required dependencies:
source scripts/install_mac.sh
- Download models' weights:
bash scripts/download_data.sh
Copy your avatars into avatars
folder. Crop pictures to make them square. Prefer pictures with uniform background.
Your web cam must be plugged-in. You can choose your camera by changing CAMID
in run*.sh
script.
Run:
bash run.sh
- Run:
bash run_mac.sh
- Go to CamTwist.
- Choose
Desktop+
and pressSelect
. - In the
Settings
sction chooseConfine to Application Window
and selectpython (avatarify)
from the drop-down menu.
Your face should fit in the red rectagle: it should not fit perfectly, but don't get too close/far. After that, press any number from 0 to 9 on your keyboard and one of the avatars from avatars
directory will appear.
Run Skype or Zoom only when Avatarify is running.
Go to Settings -> Audio & Video, choose avatarify
or CamTwist
camera.
Go to Settings -> Video and choose avatarify
or CamTwist
from Camera drop-down menu.
Our goal is to democratize deepfake avatars. To make the technology even more accessible, we have to tackle two major problems:
- Add support for more platforms (Linux and Mac are already supported).
- Optimize neural network run-time. Running network real-time on CPU is of high priority.
Please make pull requests if you have any improvements or bug-fixes.
- Zoom/Skype doesn't see
avatarify
camera. Restart Zoom/Skype and try again. - Avatar image is frozen: In Zoom, try Stop and Start Video.
- Avatrify uses First Order Motion Model for generating avatars.