The JP -> PT-BR pipeline is incredibly impressive. For me, the pipeline was always through an English intermediate step, which dilutes the nuance, but this feels direct and natural. What really blew me away was the OCR robustness, and I tested it on some grainy, low-quality scans of Hajime no Ippo, and it didn't miss a single dialogue bubble.
The AI inpainting is awesome! It cleans the original Japanese characters so effectively that the new typesetting looks like it belongs there. For someone looking to catch up on titles abandoned by local scanlators (like me), the bulk processing feature is a total lifesaver.