Faceswap Forum

That sounds about right to me. Skip Existing - Will not process any frame that exists in alignments file. Would be used for re-generating alignments for a file where "Extract Every N" was used in the past. Skip Existing Faces - Will attempt to find faces in any frame where faces were not f...

Not an AWS expert by any means, but just don't pass in the -p flag when launching faceswap. Use the -w flag to write out a preview image to the faceswap folder instead.

[Topic split from: https://forum.faceswap.dev/viewtopic.php?t=2058] I have done a not inconsiderable amount of testing around NaNs recently, really hoping I would find a solve, and I have drawn a blank. Where they get introduced in a model is inconsistent, although they nearly always seem to appear ...

The recent pushes around learn mask were due to a regression introduced with the new Phaze-A preset, so they were to fix those bugs rather than any feature adds... Re: Bisenet-face mask. An unintended consequence of full face masking was that when swapping someone with a lower hairline onto someone ...

Unfortunately that's just a message saying that Tensorflow is not OK. There's not really much I can do with it

The short answer is no. Images should all have a similar field of view, or you will run into the issues that you see. Ultimately, nothing is aligned in the 3D space in Faceswap. We do use some rough 3D modelling for different training types (i.e. legacy/face/head), but again, it makes assumptions th...

I don't really know what to add to what the message says. You have a GTX 1080 with compute capability of 6.1 Tensorcore support was not added until GPUs with at least compute capability of 7.0 (RTX 2xxx series) You don't run with mixed precision enabled, so it doesn't impact you. It is collecting mi...

The line, literally 1 line prior to your paste says:

Code: Select all

INFO     Storing Mixed Precision compatible layers. Please ignore any following warnings about using mixed precision.

Well, that sucks.

Not really sure what else to suggest, as this is not a repeated complaint.

You can try enabling "Allow Growth" under training settings, but I think it's a long shot that it will solve your issue.

Glad you got there in the end

The type of mask used should make no difference.

Please do let me know if rolling back solves your issue, so I can keep track of the problem.

Ok, downgrading Tensorflow looks like it forced a downgrade of typing extensions.

You should be able to fix the issue in exactly the same way as you did before, but instead of pip install tensorflow do:

Code: Select all

pip install typing-extensions">=4.0.0"

That's actually an unrelated bug...

Looking at your installed packages, it looks like you have downgraded successfully.

If you give me 10 mins, I can probably fix that issue.

I'm not entirely sure what you mean. If you mean that the wrong face is being swapped on (i.e. the face being swapped on is from a previous/next frame) then this is not something I've heard of before, and would suggest an issue with your video file. The process basically goes through the video frame...

No, I don't keep a changelog, as Faceswap is constantly updated.

You can view changes to code in the commit log. I try to name commits in a useful manner:
https://github.com/deepfakes/faceswap/commits/master

If you're asking how to do it, it probably won't be a good idea to do it. However, I will detail the steps to downgrade Tensorflow here. Please note, after this, you are on your own. If it goes wrong, then I am unlikely to help with troubleshooting the issue as it is not a great use of my time (if i...

For cli, not so much. The guides also work for the cli (for the most part anyway. Some tools like preview/manual etc are obviously GUI only). python faceswap py -h will give you command line parameters Hitting the Generate button in the GUI will generate the underlying command line arguments that wi...

Duplicate frames usually happen because, at some point (probably before you even received it), the video has been converted from one frame rate to another. It is very common.

Read more here:
https://helpcenter.extremereach.com/hc/ ... ate-Frames

This looks to be a regression in Tensorflow which appears to impact some users and not others. In theory, this shouldn't be able to happen, as TF allocates all of the VRAM it requires at the start, so it should not run out of VRAM several hours in to a training session. I have seen this happen once ...

This is usually because of special characters in the metadata of the video file. Find a video editing tool that you can use to strip the metadata. (Something like mp3tag should work ok for mp4 files). Alternatively, if you can provide the video file, I can try to put something in which will work aro...

Honestly, I couldn't really advise one way or another. I'm not sure how accurately Windows graph represents utilization for AI tasks. I develop Faceswap on a Xeon processor with a considerably lower clock speed than yours, but with considerably more cores. If you are just looking to upgrade because ...

Faceswap Forum

Search found 1838 matches

Re: "Skip Existing" vs. "Skip Existing Faces"

Re: AWS gpu instance error

Potential VRAM Saving techniques

Re: Latest with "Learn Mask"

Re: Error code viewing graph whilst training.

Re: Obscure close - Fish Eye Training are there any Z axis training models

Re: why am I getting a mixed_float16 error?

Re: why am I getting a mixed_float16 error?

Re: Out of Memory after 6 Hours of Steady Training

Re: Out of Memory after 6 Hours of Steady Training

Re: Out of Memory after 6 Hours of Steady Training

Re: Out of Memory after 6 Hours of Steady Training

Re: mismatch frame rate (output & original)

Re: updates

Re: Out of Memory after 6 Hours of Steady Training

Re: CLI documents

Re: mismatch frame rate (output & original)

Re: Out of Memory after 6 Hours of Steady Training

Re: get crashed my brand new computer

Re: Is my CPU a bottleneck?