Models?

[zigomi]

How does it work: does it uses pretrained models? If so, where are they? I haven't found them. Or does it make http requests to an external API?

[scribblemaniac]

The readme mentions that this application uses the PatchMatch algorithm. As I understand, it does not need training, a model, or an external API. The infill data comes from other parts of the image (or a separate image in some implementations).