Faceswap Forum

Yes, and that wont be NEAR enough.

Don't put folders in Timelapse spots that have no images in them.

The cause and effect that you describe is not correct. We never use the system Cuda and don't recommend it being installed at all. Instead we install the cudatoolkit like you said is installed. What makes you think it's not using your GPU? Is it providing an error or crashing? Or is it https://forum...

Updating Windows is unlikely to be a fix (or cause) for an issue like this. Though it may have installed the latest Visual C++ runtime.

This error is basically the same as

app.php/faqpage#f3r10

Try lowering the BS in the settings or run single process to see if it helps.

I'd need to see examples to know for sure. But Probably not.

Hypotheticals like this aren't really feasible to discuss until the card comes out and we test it. That said, if you want to talk about this sort of thing, feel free to drop by our Discord.

Faceswap is really not suitable for real time swaps. It's not what we built it for and it generally wont provide a smooth experience. That said, there is an older repo that supported it https://github.com/alew3/faceit_live . Currently you'll probably get closer to realtime results with something lik...

dicarlo11 wrote: ↑

Fri Sep 11, 2020 9:32 am

Any fix for this? I have this error everytime I try to train :/

Yes.

torzdf wrote: ↑

Tue Aug 25, 2020 9:26 am

This is generally due to special characters in filenames (although yours look ok), or a corrupt image file.

Try removing that image and seeing if the issue continues.

We're going to need more information. How did you install? The Installer? Did you follow the FAQ item? app.php/faqpage#f1r1 If it's still a problem, please provide the faceswap.log file from your installed folder or the full system information from the GUI or CLI.

You're going to have to give more information, what did you extract a video or a folder? Did you get multiple images of one face or did you just get a single image? Please give us more info about your problem.

You have to include alignments. This is no longer optional.

If it's the same people on both sides you can just do a fit train instead of a full train. See viewtopic.php?f=6&t=74 for information on how to do that.

I had this issue after update, and I was unable to train my previous model. Still I was able to use it to convertand automatically the model was converted to the new format (this model was using Dlight). Here is the error from the log file: File "Y:\faceswap2\lib\training_data.py", line 605, in _se...

If your computer shut down while training it may have corrupted the model. Check out the information on app.php/faqpage#f3r6 to restore the last known good model to see if it helps.

No, this is not possible with Faceswap. You can check out the First Order Model ( Avatarify ) for functionality like that.

Don't use WSL on Windows. Just use the Windows version of Faceswap. WSL does not recover any of the lost vram that Windows reserves so there is no reason to try to use the Linux version in WSL. (Especially docker which is exceptionally problematic)

I recommend (and the guide walks you through) using the sort FIRST then if you need to tweak some manually THEN use the manual tool.

The manual tool lets you fix the missing faces.

Don't worry about different sizes. The model only takes in set sizes anyway.