With some videos faceswap doesn't detect faces if they are too far away (even with minimum settings), too blurry or in hard angle. It's very time consuming to fix these faces with manual tool. I've done some testing with face restoration scripts with good results: First I extract faces from original frames as normal. I always use video converted to pictures as source to avoid frame problems. Then I copy frames with no faces detected to another directory and enchance them with gpen face restoration (best so far). Then I use these enchanced frames to do extraction again with original alignments file to add new alignments to it. After that when my alignments file is fixed with manual tool I extract faces from original frames to create training set. Much less work with manual tool. This could be optional feature in faceswap to enchance frames with no faces found and detect them again.
Feature request / Tip: How to improve face detecting
Forum rules
Read the FAQs and search the forum before posting a new topic.
This forum is for discussing tips and understanding the process involved for Extracting and preparing face sets for training a model in Faceswap.
If you have found a bug are having issues with the Extraction process not working, then you should post in the Extract Support forum.
Please mark any answers that fixed your problems so others can find the solutions.