Search found 1582 matches

by torzdf
Thu Jan 20, 2022 11:55 am
Forum: Training Support
Topic: Not enough VRAM after ~3 hours of training.
Replies: 1
Views: 46

Re: Not enough VRAM after ~3 hours of training.

The amount of VRAM that faceswap requires will not change over time. Most likely something else (Google chrome/game/desktop etc) is stealing VRAM which is causing FS to fail.

by torzdf
Wed Jan 19, 2022 3:42 am
Forum: Installation Support
Topic: GPU not used, "No global version found. Check Conda packages..."
Replies: 1
Views: 53

Re: GPU not used, "No global version found. Check Conda packages..."

gpu_cudu: No global version found. Check Conda package for Cuda gpu_cudnn: No global version found. Check Conda packages for Conda cuDNN No, this is correct, you don't want a global version installed unless you know what you are doing. Do some of you have a suggestion of what I should try? This &qu...
by torzdf
Wed Jan 19, 2022 3:40 am
Forum: Installation Support
Topic: Strange Installing Errorcode during installation faceswap on Win 10?
Replies: 10
Views: 336

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

You have a problem with you SSL somehow. You will need to resolve this. Ultimately FS cannot download the files it needs. As Bryan says this is a Conda (and pip) issue, but more specifically the issue is originating from your end. Could not fetch URL https://pypi.org/simple/pip/: There was a problem...
by torzdf
Wed Jan 19, 2022 3:32 am
Forum: Training Support
Topic: What does it mean when training starts wonky?
Replies: 3
Views: 112

Re: What does it mean when training starts wonky?

Yeah, really depends on the model you're using. If it's failing to start training though, you should try lowering the learning rate.

by torzdf
Wed Jan 19, 2022 3:30 am
Forum: Training Support
Topic: Very slow training performance - same hardware
Replies: 4
Views: 152

Re: Very slow training performance - same hardware

This bug has been fixed. Please delete your faceswap folder + re-run the installer.

by torzdf
Wed Jan 19, 2022 3:26 am
Forum: Training Support
Topic: Faces picture when training is complete?
Replies: 2
Views: 155

Re: Faces picture when training is complete?

This is a really hard question to answer, as there are so many factors involved (data source/variety, model used, batch size, hardware etc). However, there are examples on our Discord server. This is an example of a trained model (of probably more complexity/resolution that the one you will be train...
by torzdf
Wed Jan 19, 2022 3:23 am
Forum: Training Support
Topic: Possibility of Turning Off "Learn Mask" on an Active Training Model
Replies: 3
Views: 221

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

There is a fix for this, but I can only tell you top level, I can't take you through this step-by-step as it's involved. It is possible to create a new model which does not have "learn mask" enabled, and then port your required weights from your existing model into this new model. If you c...
by torzdf
Wed Jan 19, 2022 3:19 am
Forum: Convert Support
Topic: About colour adjustment causing face shake
Replies: 1
Views: 57

Re: About colour adjustment causing face shake

This is not something I've ever heard of, and theoretically shouldn't be possible. Seemless-clone definitely has issues and shouldn't be used.

The color adjustments just adjust the individual pixel values, no pixels are actually moved.

by torzdf
Wed Jan 19, 2022 3:18 am
Forum: Extract Discussion
Topic: (Suggestion) Implement the ability to zoom farther in on the mesh.
Replies: 2
Views: 51

Re: (Suggestion) Implement the ability to zoom farther in on the mesh.

Finally, is there somewhere I can find a template for what "constitutes" different landmark features? The mouth is often the hardest to edit because there are many points that I am not sure exactly where on the mouth each is supposed to correspond to. It seems that the "mouth" m...
by torzdf
Wed Jan 19, 2022 3:14 am
Forum: Extract Discussion
Topic: [Guide] Extraction - A Workflow
Replies: 68
Views: 470774

Re: [Guide] Extraction - A Workflow

Not sure I entirely understand the question.

If you have an extracted face (512px extracted image) and your target face is front and center, but there is also another face in the same image, but they aren't front and center then that is fine, keep it. The other face will be masked out anyway.

by torzdf
Wed Jan 19, 2022 3:11 am
Forum: Training Discussion
Topic: Should I select pics of face partially blocked?
Replies: 2
Views: 104

Re: Should I select pics of face partially blocked?

Personally, I remove obstructed faces unless I know that they have been properly masked for obstructions. If they have been properly masked for obstructions, then I would include them. Ultimately, the question is, what do you want the model to learn? A face or a microphone? It's the face, and feedin...
by torzdf
Wed Jan 19, 2022 3:10 am
Forum: Training Discussion
Topic: Will the program do it?
Replies: 1
Views: 26

Re: Will the program do it?

After quickly flicking through the video, the short answer is "yes" you could do this with Faceswap. The longer answer is it would take quite some work for some of the shots (striped lighting across the face for example).

TBH, it's not the conversion that I would attempt for my first swap.

by torzdf
Wed Jan 19, 2022 3:06 am
Forum: Training Discussion
Topic: Does adding Black and White footage to training data help or hinder?
Replies: 3
Views: 164

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

I'd put the B+W ones in, but would probably ensure they didn't take more than 15-20% of my total dataset.

This is pure "finger in the air" stuff though.

by torzdf
Wed Jan 19, 2022 3:05 am
Forum: Training Discussion
Topic: Halo effect around face on side B causes side A to have halo effect and not cover full face.
Replies: 3
Views: 713

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

He means using an external tool (e,g. After Effects/Premiere/many other video editing suites) to load in the frames and composite there. I have no experience of this, so would need to leave to others to advise.

by torzdf
Wed Jan 19, 2022 3:01 am
Forum: Training Discussion
Topic: Best settings for photos to photos
Replies: 3
Views: 200

Re: Best settings for photos to photos

Those previews should almost definitely look better than that at 650k on dfaker (assuming a reasonable batch size). As Bryan says, blurriness is almost almost down to data quality. Variety is key.

by torzdf
Wed Jan 19, 2022 2:57 am
Forum: Training Discussion
Topic: My Phaze-A Training results
Replies: 1
Views: 261

Re: My Phaze-A Training results

Are you able to post an image to show what you mean by "skewing". It's not an issue I have heard of before.

by torzdf
Wed Jan 19, 2022 2:54 am
Forum: Cloud Discussion
Topic: Google Colab Notebook
Replies: 1
Views: 63

Re: Google Colab Notebook

I don't use Colab, but if people test this notebook and feedback that it all works well, I will pin it as I have a feeling that the currently pinned notebook is now a little out of date.

by torzdf
Wed Jan 19, 2022 2:52 am
Forum: Artificial Intelligence
Topic: Thesis: Facial scar faceswap
Replies: 2
Views: 167

Re: Thesis: Facial scar faceswap

Hmmm, I'm going to take an opposing view and say I'm not sure this would work how you would want. The model would never have seen scarring for the target face, so at best it would have to imagine what the decodings it has received means, at worst it will be a total mess. Faceswapping trys to keep fa...
by torzdf
Wed Jan 19, 2022 2:45 am
Forum: Hardware
Topic: Questions on GPU compatibility and nvlink compatibility?
Replies: 1
Views: 120

Re: Questions on GPU compatibility and nvlink compatibility?

I cannot speak specifically for the card, but theoretically it should be ok.

https://developer.nvidia.com/blog/accel ... a100-gpus/