Search found 790 matches

by bryanlyon
Thu Mar 23, 2023 5:34 pm
Forum: Training Discussion
Topic: [Newbie] Train with another source
Replies: 2
Views: 1236

Re: [Newbie] Train with another source

When you train a model you're training it on two faces, not two videos. You can then use it to swap those two faces in any videos. If you are using an unbalanced model such as Unbalanced or Reface then you'll get the best results putting the B face onto the A video, other models do both directions e...
by bryanlyon
Tue Mar 21, 2023 9:27 pm
Forum: Training Support
Topic: Mask in training differ from mask in manual tool
Replies: 6
Views: 1540

Re: Mask in training differ from mask in manual tool

I can't tell for sure due to the quality of the screenshot, but it looks like you might be using the "extended" mask for training. If you didn't specify another mask that is the one that you'd be using. To change it, open the GUI then Settings -> Configure Settings -> Train -> Loss -> Mask...
by bryanlyon
Mon Mar 20, 2023 5:33 am
Forum: Extract Discussion
Topic: Masking is off when adding new faces in manual
Replies: 4
Views: 4833

Re: Masking is off when adding new faces in manual

Yes. I try to do everything one step at a time. It prevents issues such as where you end up with incorrect masks or faces that you've deleted from the alignments being used to train. It's much smoother and easier to avoid mistakes if you process everything one step at a time.

by bryanlyon
Sun Mar 19, 2023 5:12 pm
Forum: Extract Discussion
Topic: Masking is off when adding new faces in manual
Replies: 4
Views: 4833

Re: Masking is off when adding new faces in manual

My preference is to create new faces, use the alignment tool to extract a new faceset, then work on masks on their own. I don't ever mix the alignments and masks, this helps avoid any issues like this.

by bryanlyon
Thu Mar 16, 2023 11:55 pm
Forum: Convert Support
Topic: Dark spots on forehead after converting?
Replies: 2
Views: 3968

Re: Dark spots on forehead after converting?

I agree with Torzdf that it's probably a data/variety issue, but I think you might be able to mitigate the issue by experimenting with options in the Preview tool. That lets you try different options for conversion. Some that you might want to look at is the color correction options and the mask adj...
by bryanlyon
Thu Mar 16, 2023 11:53 pm
Forum: Training Support
Topic: Training freeze at random iteration - HRESULT failed with 0x887a0005
Replies: 4
Views: 2782

Re: Training freeze at random iteration - HRESULT failed with 0x887a0005

One of the suggestions at the link that Torzdf posted is to disable the timeout. The timeout is to prevent freezing your GPU. The suggested setting it to 10 seems like a good idea as it should still prevent freezing for longer than that, but still allow kernels that take longer than 2 seconds to run.
by bryanlyon
Thu Mar 16, 2023 7:07 pm
Forum: Training Discussion
Topic: Swapping long videos
Replies: 3
Views: 1680

Re: Swapping long videos

I don't recommend training with such large datasets. Not for hard reasons, just it's a huge pain to clean. I also don't recommend converting whole TV episodes all at once. Much better to do it scene-by-scene. Then you can ensure that the quality level is high for each scene before moving on. There a...
by bryanlyon
Wed Mar 01, 2023 8:53 am
Forum: Hardware
Topic: Hardware Critique Request/Recommendation
Replies: 1
Views: 3834

Re: Hardware Critique Request/Recommendation

So the biggest drawback in your system (and the most important part for any AI work) is the GPU. Your 1050 isn't bad by any means, but you're lacking tensor cores that really speed up the AI process. Upgrading your GPU to a newer card will benefit you more than it's benefit someone using it just for...
by bryanlyon
Tue Feb 21, 2023 11:14 pm
Forum: Training Support
Topic: Training speed tapers off after a two minutes
Replies: 4
Views: 1239

Re: Training speed tapers off after a two minutes

It's really hard to say. It's very possible that it's your CPU being the bottleneck, or even the PCI-E interface. When training, Faceswap has to load the images, warp them, augment them, and generally prepare them for the GPU before transferring them over. DNY256 is a high resolution model and we're...
by bryanlyon
Tue Feb 14, 2023 11:40 pm
Forum: General Discussion
Topic: New startup dedicated to FaceSwap hiring!
Replies: 0
Views: 10399

New startup dedicated to FaceSwap hiring!

So, We've now created a new Startup dedicated to building and working on FaceSwap! You can visit us at https://arkavian.com/ This new company is focused on FaceSwap and will be keeping FaceSwap's ethos of Open, easy-to-use, and ethical. We're excited to continue to improve FaceSwap and as part of th...
by bryanlyon
Tue Feb 14, 2023 7:26 pm
Forum: Training Discussion
Topic: [Guide] Introducing - Phaze-A
Replies: 89
Views: 165565

Re: [Guide] Introducing - Phaze-A

Not a problem, just trying to help you avoid edits on your posts.

by bryanlyon
Tue Feb 14, 2023 6:55 pm
Forum: Training Discussion
Topic: [Guide] Introducing - Phaze-A
Replies: 89
Views: 165565

Re: [Guide] Introducing - Phaze-A

Also, a note, you can "save draft" at the bottom of a post that lets you edit it before you decide it's ready for you to post it ;) .

by bryanlyon
Mon Feb 13, 2023 11:37 pm
Forum: Convert Support
Topic: OSError: Unable to open file (truncated file)
Replies: 4
Views: 6020

Re: OSError: Unable to open file (truncated file)

Please check that the model folder is really /Users/Local-Louis/Downloads. That is your downloads folder and it is HIGHLY unlikely to be the correct folder for your model. If it is, I suggest instead moving it to a folder of it's own properly and trying again.

by bryanlyon
Fri Feb 10, 2023 7:55 pm
Forum: Installation Support
Topic: Installatim on Linux Mint 21.1
Replies: 10
Views: 3555

Re: Installatim on Linux Mint 21.1

Looks like something failed to install. Can you (from inside the faceswap conda environment) get us your sysinfo. GUI Users: Go to Help -> Output system information CLI Users: From inside your virtual environment, inside your faceswap folder, run: python -c "from lib.sysinfo import sysinfo ; pr...
by bryanlyon
Wed Feb 08, 2023 8:15 pm
Forum: General Discussion
Topic: Adjusted masks during manual adjustment - didn't seem to work with training - error on conversion.
Replies: 21
Views: 11977

Re: Adjusted masks during manual adjustment - didn't seem to work with training - error on conversion.

I find sorting first saves more time than writing it out later takes. But I guess your mileage may vary.

by bryanlyon
Sun Feb 05, 2023 5:37 am
Forum: General Discussion
Topic: More Accurate Align Tool?
Replies: 3
Views: 4442

Re: More Accurate Align Tool?

Google's face mesh is unfortunately not publicly available. It's only usable through API calls on (some) Android devices.

by bryanlyon
Wed Feb 01, 2023 10:59 pm
Forum: Training Support
Topic: Everything says to switch to "No Warp" eventually, but when I to, it looks clearly like the source face again.
Replies: 7
Views: 2202

Re: Everything says to switch to "No Warp" eventually, but when I to, it looks clearly like the source face again.

DFl-h128 is a particularly "blendy" model. It's known to not create particularly distinct A and B. It's very possible that model really requires warp to work reliably. Unfortunately that is not a model that I have much experience using and can't provide more precise advice but I can say a ...
by bryanlyon
Thu Jan 26, 2023 7:01 pm
Forum: Training Discussion
Topic: Misalignment on face "Eyes on forehead"
Replies: 2
Views: 704

Re: Misalignment on face "Eyes on forehead"

This is perfectly normal and you can completely ignore it. Anything outside of the masked area is essentially a fever dream of the network and has no effect on the final results.

by bryanlyon
Tue Jan 24, 2023 4:23 pm
Forum: Convert Support
Topic: Do I really have to adjust every frame manually for conversion?
Replies: 2
Views: 3973

Re: Do I really have to adjust every frame manually for conversion?

You only need to adjust faces that aren't accurate or are missing. If they're fine then you move on to the next. Our face detector has the most problems with things like obstructions covering most of the face and the face turning away from the camera. So look especially for those frames to make sure...
by bryanlyon
Thu Jan 19, 2023 8:16 pm
Forum: Training Discussion
Topic: LPIPS Alex vs Squeeze Surprising Behavior
Replies: 4
Views: 1044

Re: LPIPS Alex vs Squeeze Surprising Behavior

This is funny because squeezenet was designed for low-memory environments. It's used on FPGAs and microcontrollers for example. However, it seems that this is likely due to how Tensorflow works. Accumulates happen after nearly every operation, but TF makes new ones for each operation instead of re-u...