Faceswap Forum

Use the preview tool to get a value which looks good.

There won't be a crash log for this.

Easiest solution is just don't use time lapse. If you must use timelapse, then they must be faces that exist within your training set.

I have not seen this error before. I suspect it's a false positive and is actually an indication that you are running out of VRAM.

Try lowering the batch size and see if it makes a difference.

The timelapse folder should only contain faces from your corresponding training set.

Try manually installing MiniConda into a location which does not have spaces in the path, then specifying the manual location when installing Faceswap

Uninstall your global versions of Cuda.

It only gets generated the first time you run faceswap

Personally, I've never needed to change the learning rate.

When you're getting towards the end though, it can help to turn off "no-flip", and also "disable warp" (in the trainer settings). Just remember to turn it back on again prior to starting a new train.

It depends. If the coverage is 100%, then 256 extract will be fine. If the coverage is anything below 100%, then you'll want to extract at a higher resolution (maybe 384?)

You can try different normalzation settings. Sometimes this gets more or less faces.

This is a macOS specific error. I'll put a pin in this and try to get a fix in the next time I swing by the manual tool.

Also, Enter your extract setting into Faceswap, hit the "Generate" button and post the output here.

And this:
app.php/faqpage#f0r3

To expand on this a little. Adjusting the landmarks adjusts the location of the face in the frame and therefore the face image. You can only adjust the face image if you have the full frame to work with.

Don't use the cv2-dnn aligner, it isn't very good.

Start > Anaconda Prompt

conda activate faceswap
set

Post output.