Output is black blob on face of target image -- tips for debug?

I have the inference up and running on my own machine, but when attempting to reproduce the image-only example from Colab (Musk->Beckham), the face of the source is not applied to the target, and I see a black blob on the face of the target instead (as shown below). Are there any tips available on how to debug this scenario?

Fantastic work on the tool, BTW -- the example results look great, and I'm really looking forward to getting it working.