Search found 570 matches

by bryanlyon
Wed Mar 03, 2021 4:11 pm
Forum: Extract Discussion
Topic: Manual alignment tool for CLI users
Replies: 2
Views: 20

Re: Manual alignment tool for CLI users

Yes, you can edit the alignments on a different system from training, you'll just need to download the alignments file and the original source video to your local system and then you can run the manual tool and re-upload the alignment file. Keep in mind you'll have to re-extract to get the updates f...
by bryanlyon
Wed Mar 03, 2021 4:07 pm
Forum: Convert Support
Topic: opencl_amd_ellesmere.0 error
Replies: 4
Views: 64

Re: opencl_amd_ellesmere.0 error

The other errors you were getting weren't related to the first error. Looks like you were training and had a face that had no alignments. Glad the restore fixed it for you.

by bryanlyon
Wed Mar 03, 2021 4:06 pm
Forum: Artificial Intelligence
Topic: Faceit_live3
Replies: 1
Views: 17

Re: Faceit_live3

We can only provide support for FaceSwap. Related technologies like First Order Model and Faceit_live are unsupported. You're going to get much better answers from asking the creators/users of that software.

by bryanlyon
Tue Mar 02, 2021 4:03 pm
Forum: Convert Support
Topic: opencl_amd_ellesmere.0 error
Replies: 4
Views: 64

Re: opencl_amd_ellesmere.0 error

Without the crash report I can't be sure, but this looks like the model file has corrupted. Make a copy and try the restore tool to see if that fixes it. Otherwise, post your crash log to a cloud service and share the link it here.

by bryanlyon
Tue Mar 02, 2021 3:59 pm
Forum: Training Discussion
Topic: Would someone mind taking a look at my data set?
Replies: 2
Views: 58

Re: Would someone mind taking a look at my data set?

Using one video per side is usually a way to guarantee poor quality. There simply isn't enough variety in either side of this dataset. You need to focus on a large number of different lighting conditions, expressions, and poses. In your case you have some variety in expression and pose, but none in ...
by bryanlyon
Tue Mar 02, 2021 3:54 pm
Forum: Installation Support
Topic: Installing on a Mac without using sudo
Replies: 1
Views: 47

Re: Installing on a Mac without using sudo

Yes, it's safe as long as you follow the directions it wont be able to damage your system. Using sudo would be the only way to damage the OS, and as long as you avoid that, your system can protect itself. Just follow the directions and don't add sudo. The explanation in the instructions explicitly s...
by bryanlyon
Tue Mar 02, 2021 3:43 am
Forum: Convert Support
Topic: Converted File Faces Haven't Changed
Replies: 4
Views: 76

Re: Converted File Faces Haven't Changed

Since you say you already tried the swap model, try clicking the "generate" button and post the result in here.

by bryanlyon
Tue Mar 02, 2021 12:02 am
Forum: Installation Support
Topic: Mac OS High Sierra installing problem
Replies: 2
Views: 61

Re: Mac OS High Sierra installing problem

Another issue, "tensorflow-gpu" is made for Nvidia GPUs. Macs do not support Nvidia GPUs. Instead you'll install tensorflow (non -gpu) and plaidML to use an AMD GPU. The proper process to do that is in the install guide that torzdf posted.

by bryanlyon
Mon Mar 01, 2021 11:50 pm
Forum: Training Support
Topic: Mask problem
Topic: Mask problem
Replies: 10
Views: 165

Re: Mask problem

Updating your files to add masks is quite easy and does not require re-extracting. You can simply use the mask tool to add the missing masks.

by bryanlyon
Mon Mar 01, 2021 8:08 pm
Forum: Hardware
Topic: Quadro K2000M Problem
Replies: 2
Views: 46

Re: Quadro K2000M Problem

Faceswap seems to be seeing your GPU, however, you do have only 2gb of vram. That means that some tasks wont be possible on that GPU (including probably training). Unfortunately, we really recommend larger GPUs since you'll struggle getting anything usable out of that GPU.

by bryanlyon
Fri Feb 26, 2021 5:11 pm
Forum: Training Discussion
Topic: A couple small questions
Replies: 1
Views: 67

Re: A couple small questions

Most likely, the data you added in for B side had more variety than the data for A. Low variety is easier to recreate and so has lower loss.

by bryanlyon
Mon Feb 22, 2021 4:56 pm
Forum: Training Discussion
Topic: "Disable Warp", not recommended but mandatory
Replies: 5
Views: 331

Re: "Disable Warp", not recommended but mandatory

No, turning it off and on is actually just recreating what we already do. Not every image is warped, but at the end, turning all warping off help reduce the warped effects on the convert.

by bryanlyon
Thu Feb 18, 2021 9:39 pm
Forum: Artificial Intelligence
Topic: How to deepfake, but with input expression kept. (like snapchat)
Replies: 2
Views: 167

Re: How to deepfake, but with input expression kept. (like snapchat)

If it's for non-commercial uses, you can check out https://github.com/dunnousername/yanderifier which uses the First Order Motion model to transfer motion from one face to another. This doesn't match faceswap for quality, but provides a quick option.

by bryanlyon
Sat Feb 13, 2021 1:16 am
Forum: Installation Support
Topic: !!! DO NOT UPDATE NVIDIA DRIVERS TO 460.79 !!!
Replies: 1
Views: 3545

Re: !!! DO NOT UPDATE NVIDIA DRIVERS TO 460.79 !!!

This has since been solved with the >461 drivers. If you run into this problem, upgrade your drivers to the latest from Nvidia.

by bryanlyon
Sat Feb 13, 2021 1:14 am
Forum: General Discussion
Topic: Read Me! Frequently Asked Questions
Replies: 0
Views: 719

Read Me! Frequently Asked Questions

We have an FAQ at https://forum.faceswap.dev/app.php/faqpage (also linked in the bar up top next to Rules). Most of your questions are answered there, and you can save a lot of time if you read through them before asking a question. If you do ask a question that's answered in the FAQs, don't be surp...
by bryanlyon
Sat Feb 13, 2021 1:06 am
Forum: Training Support
Topic: unable to train and convert or preview model
Replies: 2
Views: 119

Re: unable to train and convert or preview model

Can't be sure since you didn't post a crashlog, but that last message looks like a corrupt model file. If so, you can try to use the restore tool to go back to a previous version. Please post the crashlog if that's not it. Your first error is a GUI glitch that happens (particularly when the model is...
by bryanlyon
Fri Feb 12, 2021 6:43 pm
Forum: General Discussion
Topic: Challenges faceswapping VR videos
Replies: 7
Views: 209

Re: Why you shouldn't Faceswap VR videos

Yes, you can just add that, not to the end, but before the output video gets declared. Best advice I can give is to experiment and learn how ffmpeg works. It's really quite powerful.

by bryanlyon
Fri Feb 12, 2021 12:02 am
Forum: Extract Discussion
Topic: Manual Alignments Tool issues
Replies: 5
Views: 184

Re: Using Manual Tool with folder of images as source

You can use images, but don't give it the folder of extracted faces, you need to give it the original folder you extracted from.

by bryanlyon
Wed Feb 10, 2021 9:32 pm
Forum: General Discussion
Topic: Challenges faceswapping VR videos
Replies: 7
Views: 209

Re: Why you shouldn't Faceswap VR videos

The effmpeg frontend in FaceSwap is just a tiny slice of what ffmpeg can do natively. You'll have to learn to use the CLI to unlock it's full power.

A basic use of ffmpeg to crop is:

Code: Select all

ffmpeg -i in.mp4 -filter:v "crop=out_w:out_h:x:y" out.mp4
by bryanlyon
Wed Feb 10, 2021 9:12 pm
Forum: General Discussion
Topic: Challenges faceswapping VR videos
Replies: 7
Views: 209

Re: Why you shouldn't Faceswap VR videos

The solution to the "lost detail" is actually pretty simple. You just set the output of the dewarp to a much higher resolution to account for the dewarped resolution. Normally, warped images have much higher pixel density at the center since that's where the warp is the least. You just nee...