This is no way near enough information.
Please post the direct output of that command, and also a screengrab of your ~/faceswap
folder
This is no way near enough information.
Please post the direct output of that command, and also a screengrab of your ~/faceswap
folder
Masking out the mic in the alignments file for the video you are converting (with the manual tool) should suffice. Just make sure you are using the correct mask.
I'm pretty certain it doesn't work, and you have to use the -w
flag to write out the preview image to the faceswap folder (which you can then pull down locally to view)
This is most likely hardware/psu related.
With out a full system output there is little more to add.
You will need to start a new model each time you lower the learning rate. You won't be able to pick up and continue on an already corrupted model.
Lower the learning rate.
Generally, more complex models will need a lower learning rate. Dlight can be a bit unstable at times.
Yeah, this happens when deleting faces. Unfortunately it's a very hard bug to track down, so I haven't prioritized.
You can still delete faces without the landmarks, and a reload of the tool will bring everything back in line.
I tend to keep any in which crop above the eyes, but discard others. As long as there aren't too many of them, it shouldn't matter too much.
There have been no changes to Extraction code for a while, and I have never seen this.... not that this is much help to you, I'm afraid. I'm not even sure how to start diagnosing on that
It looks like all your images are failing to load. Try using paths without any special characters in their name (e.g. é). Specifically there may be an issue with the folder name téléchargement
It should. I will need to check. I will look when I have a chance (may not be for a while).
Thanks for the heads up. I'm tagging your post with bug
to remind me to revisit.