Faceswap Forum

The installer just downloads and installs from the github. If it's failing on conda then you might want to manually install miniconda and provide the path to the faceswap installer.

You wont find a fix for this in the model settings. In the case of vastly different face shape you're going to need to do manual compositing to get good results. To do that, change the "writer" to OpenCV or Pillow and enable "draw Transparent" to get just the model output. You ca...

Yeah, just use the augmentation setting. That's it's purpose. And while the slider only goes up to 25 degrees, it'll actually work with any number manually inserted.

That's really to complicated to answer. Depending on your data available, it might be that you NEED the extra data to get reasonable results. It could also equally be that you would just get better results from your color data. Unfortunately there is no one "correct" answer without having ...

This is answered in our FAQ.

app.php/faqpage#f1r1

The zoomed in version is accurate.

All required alignments for training are included in the training images. The .fsa is now used only for final conversion.

The mask you're using isn't removing the extra hair, I suggest using BiSeNet with the hair disabled to remove that.

No. Latency is bad enough if the GPUs are all in one box. If you were to sync over a network you'd lose anything you could possibly gain from more GPUs.

There are no maximum number of GPUs, but each additional GPU adds more latency and provides less gain. If you were to do 8 GPUs, you'll definitely actually be past that bottleneck and you'll get better results by reducing the number of GPUs. The GPU I normally recommend is a 3060. It's got a lot of ...

Should be feasible. Though you'll need to take resolution limits into account. Makeup has been done in the past successfully.

Learn Mask changes the model itself in order to let it learn the mask. Unfortunately once a model is started, it can't be changed. You're going to be stuck with the mask being learned, that said, once convert comes up, you can just not use the predicted mask.

This chain of questions sounds like an XY problem ( https://en.wikipedia.org/wiki/XY_problem ). Why don't you ask us for help with what you're actually attempting to do. If you're just trying to modify the extract process of Faceswap for other extract needs, you're going to have to modify everything...

Sure, 2080 ti is a fine card. You wont be able to go as big but in many tasks your card will run faster. It's a balancing act with the current cards. Especially when you consder the lack of availability in the retail channels. In general, unless you have specific needs, I wouldn't recommend updating...

You shouldn't modify the image at all. We include metadata in the image that is required to train. Any modification to the image stands the chance of losing that data. And things such as rotating will invalidate the metadata making it incorrect and unusuable. This sounds like an X-Y problem ( https:...

Dlight is prone to NAN errors. It's just a part of it's design that it's more likely to hit a NAN. You can try reducing the Learning Rate to reduce the chance of NANs. There are also other things you might try to reduce them, but you wont completely eliminate them. When you hit a NAN the recommended...

3060 is my recommended card. See viewtopic.php?f=16&t=10

That said, right now it's the card you can get more than the card you want.

Short scenes let you focus your efforts. Even if you're doing large amounts of scenes, I recommend converting one at a time. It lets you tweak settings and efforts to get the best results for each situation.

Generally you'll only want to fine-tune data that you're using for actual conversion. It's not worth fine-tuning thousands of files for training. Better to just delete bad images rather than tuning them.

Like it says, to use the faster clustering with that many images you'd need more than 700gb of RAM. It uses exponential amounts of RAM to cluster based on how many images you have. You can split it to subses of 50k images to sort it in 4 packs if you want to use the faster method.

Faceswap Forum

Search found 790 matches

Re: Strange Installing Errorcode during installation faceswap on Win 10?

Re: Halo effect around face on side B causes side A to have halo effect and not cover full face.

Re: Training on rotated Images

Re: Does adding Black and White footage to training data help or hinder?

Re: Strange Installing Errorcode during installation faceswap on Win 10?

Re: The mask appears to be covered in two ways.

Re: How to improve training?

Re: Newbie question, artefacts on converting

Re: question on the use of multiple gpus 8+

Re: question on the use of multiple gpus 8+

Re: Thesis: Facial scar faceswap

Re: Possibility of Turning Off "Learn Mask" on an Active Training Model

Re: How to remove the padding from the extracted faces

Re: GPU RAM Size vs. TFLOPS: which matters more?

Re: Training on rotated Images

Re: CRITICAL NaN Detected and cannot train

Re: GPU RAM Size vs. TFLOPS: which matters more?

Re: Manually adjusting thousands of frames individually?

Re: Manually adjusting thousands of frames individually?

Re: Not enough RAM to perform linkage clustering. Using vector clustering. This will be significantly slower.