Search found 22 matches

by deephomage
Sun Apr 05, 2020 1:12 am
Forum: Extract
Topic: Stuck at initializing S3FD
Replies: 2
Views: 11

Re: Stuck at initializing S3FD

You haven't provided enough information to troubleshoot the problem. Did you use the Windows installer?
by deephomage
Wed Mar 25, 2020 12:43 am
Forum: Hardware
Topic: Dual p106-6gb
Replies: 11
Views: 149

Re: Dual p106-6gb

The entire crash log is needed for troubleshooting. Please post it [Editted]Here, or[/edit] in the appropriate discord channel.
by deephomage
Tue Mar 24, 2020 8:25 pm
Forum: Extract
Topic: Minimizing Zoom/Black space
Replies: 2
Views: 49

Re: Minimizing Zoom/Black space

You can minimize the number of false-positive face detections by setting a minimum detection size in pixels using the slider in the extract tab in the GUI. (e.g. 100 or more). But use the minimum size option with caution and check your results carefully. If you're having issues with "zoomed-out" fac...
by deephomage
Mon Mar 23, 2020 8:40 pm
Forum: Hardware
Topic: Dual p106-6gb
Replies: 11
Views: 149

Re: Dual p90-6gb

Are you sure your motherboard can support three graphics cards? Hard lockups usually indicate an IRQ/hardware conflict. If you're not still mining, the simplest solution is to sell the mining cards and use the proceeds to buy an Nvidia 10x or 20x series card with 6 Gb VRAM or more. Hardware advice i...
by deephomage
Sat Mar 21, 2020 7:02 pm
Forum: Training
Topic: Train Rate between Two Sessions
Replies: 2
Views: 73

Re: Train Rate between Two Sessions

If you mean Learning Rate, yes. It's configurable under Training Settings, Global.
by deephomage
Wed Mar 11, 2020 2:23 am
Forum: Extract
Topic: odd Training error
Replies: 13
Views: 324

Re: odd Training error

The error message tells you what you should do (near the end). You should run the Mask Tool to generate the Unet-DFL mask for your faceset, or select a different mask in the training configuration options. Look under Tools, Mask and select the Unet-DFL option. Provide the alignments.fsa location and...
by deephomage
Wed Mar 04, 2020 1:29 pm
Forum: Training
Topic: question: I Ttride to do like a custom LowMem plugin / Model.py
Replies: 1
Views: 232

Re: question: I Ttride to do like a custom LowMem plugin / Model.py

Read the training guide: https://forum.faceswap.dev/viewtopic.php?f=6&t=146. It's not recommended that you edit the model files or reduce the autoencoder dims. The Original model has a lowmem option for VRAM limited GPUs. Use the lowmem option. Also, try reducing your batch size and closing all othe...
by deephomage
Sun Mar 01, 2020 5:55 pm
Forum: Convert
Topic: Conversion problem
Replies: 3
Views: 412

Re: Conversion problem

I would suggest that you use a 1 to 2-minute destination clip, not attempt to convert 80,000 frames. If only the fifth frame of the destination clip is converted, it sounds like something went awry during the extraction process. You need to extract every face in the destination clip and remove the u...
by deephomage
Sat Feb 22, 2020 3:39 pm
Forum: Extract
Topic: It's possible to change face landmarks after extract?
Replies: 1
Views: 305

Re: It's possible to change face landmarks after extract?

Yes, it's possible to fix the face landmarks after extracting. Look under Tools, Alignments, and select the Manual option. You can then edit the landmarks on a frame by frame basis and correct misalignments. After manually editing the alignments, you need to use the Extract option under Tools, Align...
by deephomage
Mon Feb 17, 2020 12:45 am
Forum: Convert
Topic: conversion fails, missing alignments
Replies: 3
Views: 569

Re: conversion fails, missing alignments

You need an alignments file for the faces in the destination clip you want to use for the swap. If you're confused as to which alignments file corresponds to the destination clip you want to use, then the easiest solution is to re-extract the faces the from destination clip you want to use and save ...
by deephomage
Sun Feb 16, 2020 5:24 pm
Forum: Training
Topic: Very low EGs training
Replies: 2
Views: 314

Re: Very low EGs training

Your EGs are low because you're training with the CPU version of tensorflow rather than the GPU version, tensorflow-gpu. In your faceswap environment:

conda install tensorflow-gpu==1.15

Alternatively, delete the faceswap folder and run the installer again and choose Nvidia.
by deephomage
Fri Feb 14, 2020 11:07 pm
Forum: Training
Topic: Training stopped without error message
Replies: 5
Views: 285

Re: Training stopped without error message

Docker isn't officially supported. You're welcome to try Windows, Linux or macOS,
by deephomage
Fri Feb 14, 2020 11:03 pm
Forum: Convert
Topic: I have a conversion problem
Replies: 16
Views: 1353

Re: I have a conversion problem

I didn't create the Witcher deepfakes, you' ll have to ask the creator. Generally, it takes at least a few days to train a model on a pair of faces. The higher the resolution, the longer it takes to train the model. I use an RTX 2080ti, but an RTX 2060 Super, or a used GTX 10 series card with 8 Gb. ...
by deephomage
Wed Feb 12, 2020 8:07 pm
Forum: Extract
Topic: [Guide] Extraction - A Workflow
Replies: 43
Views: 56487

Re: [Guide] Extraction - A Workflow

In a broad sense, a clean set of images means keeping the extracted faces you want to swap, deleting unwanted faces extracted from the video clip and also deleting any extracted objects falsely identified as faces.
by deephomage
Wed Feb 12, 2020 3:31 pm
Forum: Convert
Topic: I have a conversion problem
Replies: 16
Views: 1353

Re: I have a conversion problem

I don't know what guide you're referring to, but your processor and video card will not "burn out" or get damaged by training a model. You do need to have good cooling for your CPU and video card because training a model stresses your system and takes several days of 24/7 computer time. 720 or 1080 ...
by deephomage
Tue Feb 11, 2020 10:03 pm
Forum: Convert
Topic: I have a conversion problem
Replies: 16
Views: 1353

Re: I have a conversion problem

The Faceswap forum really isn't the place to find answers about conversion issues with DeepFaceLab. You should post your problems with the software on the DFL developer's github page. Sometimes other DFL users will offer advice. You could also try contacting the creator of the video you posted and a...
by deephomage
Thu Jan 23, 2020 6:47 pm
Forum: Extract
Topic: [Guide] Extraction - A Workflow
Replies: 43
Views: 56487

Re: [Guide] Extraction - A Workflow

To train a model, you need to extract both pairs of faces from video clips, to make a face set. For example, to swap Keanu Reeves and Nic Cage, you would extract Keanu and Nic faces from a variety of sources, to make a training set for each. If the video clips also contain unwanted faces, just delet...
by deephomage
Thu Jan 09, 2020 7:41 pm
Forum: Installation
Topic: [Tip] How to Update Faceswap to the Latest Version
Replies: 2
Views: 1654

Re: [Tip] How to Update Faceswap to the Latest Version

Yes, the code has changed since that post was written, I'll correct the original post, thanks.
by deephomage
Thu Jan 09, 2020 4:51 pm
Forum: Training
Topic: How to record a video for the sole purpose of training/face swapping?
Replies: 6
Views: 260

Re: How to record a video for the sole purpose of training/face swapping?

I've done deepfake work for clients who shot their own footage, and my suggestions would be: Aim for about 15-20 minutes of high-quality video footage of each person's face to obtain enough images to make a face set for each person. Medium distance, front face views, as in an interview, work best. T...
by deephomage
Mon Nov 25, 2019 3:56 pm
Forum: General Chat
Topic: Concentrate training on only "bad" Images
Replies: 2
Views: 472

Re: Concentrate training on only "bad" Images

I'm not sure why you would want the network to train on "bad" images. If you're seeing poor quality in certain images in the previews, you should manually align to check the alignments for those images. Poor quality face recreations may also point to a lack of image quality or not enough diversity o...