Here is a test done in Blender joining audio2face and deeplabcut.
The first loop is only with audio2face, and the second loop is audio2face + deeplabcut.
This test was done in less than 1h, so you can see that when deeplabcut enters, it get very distorted. Its because I still have to fix the amount of movement and the weightpaint.
This will be lots of work, maybe in some weeks we will have more stable version to show.