Faceswap Forum

Oh god, that site sounds HORRIBLE.

See

viewtopic.php?f=6&t=74

This is a memory related error, yes. 4GB of system RAM is really just not enough and is causing problems in this case. You can try switching to a SIngle job (set jobs to 1) and see if that helps, but I think you're likely to need an upgrade to your system RAM before you'll be able to do much with Fa...

It's there, try reading from the start.

Yes, we welcome any guides you want to make. People keep asking for a video guide if that sounds doable, we'd appreciate it. Even if it's text, it may be helpful.

See the hardware guide at viewtopic.php?f=16&t=10 .

When I say use the benchmark, I just mean use it for a short time while you start the training process, but you should still use both for Faceswap. It's just there to "clear the hurdle" of the demand spike. You'll probably have an easier time of lowering the power needs.

This is probably due to drivers/Windows more than the card. Quadro cards use different drivers and probably reserved less memory for OS use, leaving more memory for Faceswap. Windows will often reserve 1/4-1/2 your total vram. 4gb is really close to the minimum we support, and it's likely that you'r...

Chances are this is due to a slow response by your PSU to the sudden demand of 2 GPUs. This is probably creating a "brown out" type situation where due to a temporary voltage drop your system hangs. Faceswap (like most GPU compute tasks) can cause a sudden spike in voltage draw when first starting. ...

Yes, you need to extract any videos you want to train with. It's highly recommended you use more than 2 videos. See viewtopic.php?f=5&t=27

Best way to avoid this may be to extract the video into the frames and go from there. Sadly there are all sorts of frame issues in videos which show up in different contexts. Extracting to PNGs is sometimes the only guarantee, but it's not perfect since recreating that video can be a pain.

Maybe the devs can offer an option of adding IFTTT integration... Not really likely to happen. This would be extremely fiddly and cause issues for some workcases. You're going to do best finding a soltuion on your own. Something like keeping track of Python processes or using "generate" and adding ...

Glad you've got a work-around. This is unlikely to be directly fixed. The manual tool is currently under a complete rework so the current tool will not likely see any fixes or changes as the new one should be here soon and will fix issues like this.

Link was broken, I've fixed it, try again.

See viewtopic.php?f=6&t=74 especially the section titled "Avoid beards that go outside the face area".

You can put it on Google Drive/Dropbox or similar. Feel free to PM a link if you don't want it public.

Every time the mask tool runs, it creates a backup of the old alignments file. Make sure you're using the correct ones for all the tasks. Make sure that none of the processes fail or throw an error (make sure that the mask tool actually creates a new FSA and it should be larger than the previous one...

Try again, chances are it errored out the first time and didn't actually generate the masks.

For Dlight, this is an intended "feature" of the model. It saves memory on the A decoder in order to maximize the quality of the B side. This also applies to Realface and Unbalanced. All the other models you mentioned are symmetrical where the A and B sides are identical and you can swap both direct...

On your questions: 1: Convergence isn't where the face is blurry, but where the model has learned all that it could. Either because you've hit the limit of your data, or the model. The face could still be blurry. 2: Original will probably converge in a couple days, and just not learn more. Villain c...