Search found 619 matches

by bryanlyon
Thu Jan 21, 2021 6:48 pm
Forum: Artificial Intelligence
Topic: Reface analogue
Replies: 5
Views: 4020

Re: Reface analogue

This is nothing like what Faceswap is doing. Faceswap is "learning" the faces and an encoding that corresponds to the input pose, expression, and lighting (though still technically not as a 3d object). Thats what the training is, building an understanding of what each face "is" a...
by bryanlyon
Thu Jan 21, 2021 4:59 pm
Forum: Installation Support
Topic: win pro, nvidia 1050ti - GUI does not launch
Replies: 4
Views: 1956

Re: win pro, nvidia 1050ti - GUI does not launch

Glad it's working, know that the best recommendation seems to be to start the process that's going to use the GPU before you enable RDP.

by bryanlyon
Thu Jan 21, 2021 3:54 pm
Forum: Artificial Intelligence
Topic: Reface analogue
Replies: 5
Views: 4020

Re: Reface analogue

Pretty sure that Reface doesn't do anything in 3d at all. They using a morphing/latent space search for their face swapping. Basically they take the features of each face and morph them together and put them back.

by bryanlyon
Thu Jan 21, 2021 1:52 am
Forum: Extract Support
Topic: Crash while sorting faces
Replies: 4
Views: 3535

Re: Crash while sorting faces

To me, this looks like a corrupted sort model. You may need to delete the Facenet_inception_vggface2_v1.h5 model file from the ".cache" folder and let it download it again. Or it could be a glitch and if you try again it works fine.

by bryanlyon
Thu Jan 21, 2021 1:45 am
Forum: Extract Support
Topic: Issue Extracting..
Replies: 2
Views: 2316

Re: Issue Extracting..

First I'm going to point out that 30xx cards currently have issues. See viewtopic.php?f=14&t=1043

If you've already tried the guide in that thread, please post any errors you have in that thread instead of in a new message.

by bryanlyon
Thu Jan 21, 2021 1:41 am
Forum: Convert Support
Topic: Converted Clip is about Half Length
Replies: 4
Views: 3574

Re: Converted Clip is about Half Length

Are there any errors in the logs or messages in the bottom half of the faceswap gui?

by bryanlyon
Mon Jan 18, 2021 10:59 pm
Forum: Training Discussion
Topic: Would a B&W data set (on both sides) train faster than color?
Replies: 6
Views: 2610

Re: Would a B&W data set (on both sides) train faster than color?

Yes, and no. Even with B&W sources, our model expects color right now. Using B&W images may save a little bit of time training right now (probably negligibly however). Better would be to modify the model (and training code) to expect 1 channel images in order to give faster results.

by bryanlyon
Fri Jan 15, 2021 5:35 pm
Forum: General Discussion
Topic: Any chance we can get a changelog for updates?
Replies: 2
Views: 3303

Re: Any chance we can get a changelog for updates?

Since we don't do "releases" in a traditional sense, but instead maintain a rolling release, we can't provide "changelogs" in the normal sense. That said, you can see a list of all changes in the github repo at https://github.com/deepfakes/faceswap/commits/master or https://githu...
by bryanlyon
Fri Jan 15, 2021 5:33 pm
Forum: Training Discussion
Topic: [Guide] Training best practices.
Replies: 11
Views: 36230

Re: [Guide] Training best practices.

Generally, it's not worth it. Much better to get an actor who matches closer. If that's simply not possible, however, you may be able to do it, as posted in the main post, get lots of data and leave color augmentation on.

by bryanlyon
Fri Jan 15, 2021 5:29 pm
Forum: Training Support
Topic: Alignments file does not exist
Replies: 15
Views: 3468

Re: Alignments file does not exist [For Old Dataset]

All training requires an alignment file. This is not optional under any settings anymore. I'm afraid you're going to have to re-extract.

by bryanlyon
Fri Jan 15, 2021 5:26 pm
Forum: Hardware
Topic: Hardware best practices
Replies: 58
Views: 42615

Re: Hardware best practices

RTX Titan has no real advantages over a 2080 ti which shares a nearly identical chip. It also barely has an advantage against a 2060 ti. Thats why it's not a recommended card. If you buy 2 of them you'll get training about 2x a 2060 ti at 10x the price. If you really want that much power, just buy 2...
by bryanlyon
Fri Jan 15, 2021 12:35 am
Forum: Hardware
Topic: Benchmarking
Replies: 3
Views: 3354

Re: Benchmarking

The problem is it's SO subjective, it varies heavily on CPU, GPU, Model, PCI-E Lanes, and more. It's almost impossible to properly acccount for them in any meaningful way.

by bryanlyon
Thu Jan 14, 2021 6:25 pm
Forum: Extract Discussion
Topic: Extracting zoomed out frames? Faces too small per frame
Replies: 8
Views: 1861

Re: Extracting zoomed out frames? Faces too small per frame

Yes, this is deliberate. See the link that Torzdf posted as to why.

by bryanlyon
Thu Jan 14, 2021 3:55 pm
Forum: Hardware
Topic: Hardware best practices
Replies: 58
Views: 42615

Re: Hardware best practices

Hi, I'm not sure if this is the correct thread, but this is hardware related. I see people in other threads getting 300 EG/s and the like with a 2070 and whatnot. Using a GTX1060 6GB with Realface seems to net an average of 35EG/s. Obviously I'm not expecting anything close to 2070 speeds, but is 3...
by bryanlyon
Tue Jan 12, 2021 4:46 pm
Forum: Training Support
Topic: Can't resume training after merging allignment files
Replies: 1
Views: 469

Re: Can't resume training after merging allignment files

domihq wrote: Tue Jan 12, 2021 4:32 pm

01/12/2021 17:25:36 ERROR Faces missing alignments for side A: ['00225.png', '00248.png']



Easiest way, delete these files.

by bryanlyon
Tue Jan 12, 2021 4:38 pm
Forum: Extract Discussion
Topic: Extracting zoomed out frames? Faces too small per frame
Replies: 8
Views: 1861

Re: Extracting zoomed out frames? Faces too small per frame

If they look like the samples given in those links, then your friend's extract is correct. If they're more zoomed out than that, please post some samples so we can see what you mean.

by bryanlyon
Mon Jan 11, 2021 5:55 pm
Forum: Extract Discussion
Topic: Extracting zoomed out frames? Faces too small per frame
Replies: 8
Views: 1861

Re: Extracting zoomed out frames? Faces too small per frame

This is expected operation. See https://github.com/deepfakes/faceswap/c ... 9868b0b662

You're running an old version while he has the current version is all.