An extension of the NerdyRodent's VQGAN-CLIP, which started out as a Katherine Crowson VQGAN+CLIP-derived Google colab notebook. The idea is to implement patch-based training technique to do a fast video style transfer using VQGAN-CLIP.