DEEPFAKE Tutorial: A Beginners Guide (using DeepFace Lab)



– Learn how to deepfake or faceswap using DeepFace Lab in this step-by-step tutorial guide for beginners. Bonus tips and tricks from a professional deepfake artist.

Watch our Deepfake Christmas Special

Follow VFXChris Ume

Download DeepFace Lab

Face Swap (Alternative)
►…

45 Comments

  1. i cant figure out were to download
    also will a 960m b good enough?

  2. I'm stuck on 5:22 I click on it and it tells me "Using plaidml.keras.backend backend
    INFO: Opening device "opencl_amd_kalindi.0"
    And that's it. I've left there for hours and it hasn't changed. What do I do? (I'm using the OpenCL version)

  3. Thos effect is not new.
    i remenber that band "everclear" already used that in the vid AM RADIO .

    Everclear – AM radio

    So cool!!!

  4. man i didnt realise how much work a deepfake is! props to people have the patience to do this!

  5. I can't even finish step 2

    Traceback (most recent call last):

    File "C:UsersMEDownloadsDeepFaceLab_OpenCL_internalDeepFaceLabmain.py", line 11, in <module>

    from interact import interact as io

    File "C:UsersMEDownloadsDeepFaceLab_OpenCL_internalDeepFaceLabinteract__init__.py", line 1, in <module>

    from .interact import interact

    File "C:UsersMEDownloadsDeepFaceLab_OpenCL_internalDeepFaceLabinteractinteract.py", line 7, in <module>

    import cv2

    File "C:UsersMEDownloadsDeepFaceLab_OpenCL_internalpython-3.6.8libsite-packagescv2__init__.py", line 3, in <module>

    from .cv2 import *

    ImportError: DLL load failed: Error performing inpage operation.

    Press any key to continue . . .

  6. I'm assuming your computer cost around £4000??? Wow the graphic's card alone is £1,149.00 :O
    Even my powerful gaming laptop, gets very hot with Deepfake and I have 6GB RAM video RAM and 16GB RAM and whilst it doesn't shut off or anything… i'd be very cautious living it alone for days rendering.

    Realistically probably 90% of users simply won't be able to do this without a sooped up computer like yourself, as I see a lot more verified accounts doing this than non.

    I've seen some great deepfakes, but most vary in quality, from blurry face replacements to distortions, which might be the number of renders.

    Of course you really need two computers if thinking about doing this, especially if you have to do a lot of daily on your computer.

  7. I was really pumped to do this, but I ran into a slew of roadblocks and it doesn't look like there's enough information out there to get me where I need to go.

    I ran into a problem both in the source video extraction and the destination where it reads there's an I/O error and cannot finish extracting all of the frames. This happened to both videos.
    I tried downloading multiple versions of the software and I have ran into this problem each time. Any idea how to fix this?

  8. Now we know how Bill
    Hader shift his face. They use the same Technology in order to work for a famous actor's. Now everything makes since now. For a minute I thought he might be a raptialian shifting!

  9. Very good tutorial, helped a lot, thanks. 9:06 " […/cuda_driver.cc:806] failed to allocate 514.81M (..) from device: CUDA_ERROR_OUT_OF_MEMORY: out of memory" Should we expect this often? How many times have you seen this error?

  10. Why can they not show us the AE part? ;( Man it seems to make such a huge difference and they don't define how to do it all, anybody know? Please help!

  11. GALERA ACABEI DE POSTAR UM VIDEO E QUERIA A AJUDA DE VOCES !! SEI QUE POUCOS IRAO LER MAS SE VC LEU, QUE DEUS TE ABENÇOE!!

  12. Hi, Great video!! I subscribe your channel for further updates! Just wanna know, is there a way to pause the training and resume again whenever you want??? So i can use my PC when it is non used by other and it is available for such task

  13. awesome, now imma make some porn videos of my crush getting rammed so I lose all my feelings for her.

  14. I don't get how batch size works, I get it's based on my GPU capabilities, but what does that translates to? For instance, I have an NVIDIA GTX 1060. What size would be recomended for my card? The thing is, I see many people talking about numbers between 1 and 8, but you put in 20(Understandable with your MSI computer) but would my PC handle a batch size that big?

  15. So the current version of this app has slightly different batch options and names and I have been able to make it all the way to step 6. Neither the SAEHD or the Quick96 training batch files work for me. The Quick96 gives me a out of memory error message that fails to allocate space from CUDA. I have 2 GPU's in my system and both generate the same error message. One is a 2080Ti and the other is a 1080Ti. The SAEHD training batch gives me a slew of other errors that I simply have no idea what they mean. I'll paste below. Any help will be greatly appreciated.

    Error: Dimensions must be equal, but are 42 and 40 for 'Conv2D_1' (op: 'Conv2D') with input shapes: [?,42,68,68], [5,5,40,84].

    Traceback (most recent call last):

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonframeworkops.py", line 1659, in _create_c_op

    c_op = c_api.TF_FinishOperation(op_desc)

    tensorflow.python.framework.errors_impl.InvalidArgumentError: Dimensions must be equal, but are 42 and 40 for 'Conv2D_1' (op: 'Conv2D') with input shapes: [?,42,68,68], [5,5,40,84].

    During handling of the above exception, another exception occurred:

    Traceback (most recent call last):

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabmainscriptsTrainer.py", line 55, in trainerThread

    debug=debug,

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabmodelsModelBase.py", line 178, in _init_

    self.on_initialize()

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabmodelsModel_SAEHDModel.py", line 177, in on_initialize

    encoder_out_ch = self.encoder.compute_output_channels ( (nn.floatx, bgr_shape))

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasmodelsModelBase.py", line 143, in compute_output_channels

    shape = self.compute_output_shape(shapes)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasmodelsModelBase.py", line 130, in compute_output_shape

    result = self.__call__(phs[0] if not_list else phs)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasmodelsModelBase.py", line 113, in _call_

    return self.forward(*args, **kwargs)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasarchisDeepFakeArchi.py", line 114, in forward

    x = nn.flatten(self.down1(inp))

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasmodelsModelBase.py", line 113, in _call_

    return self.forward(*args, **kwargs)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasarchisDeepFakeArchi.py", line 57, in forward

    x = down(x)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasmodelsModelBase.py", line 113, in _call_

    return self.forward(*args, **kwargs)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcorelerasarchisDeepFakeArchi.py", line 34, in forward

    x = self.conv1(x)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcoreleraslayersLayerBase.py", line 14, in _call_

    return self.forward(*args, **kwargs)

    File "D:DeepFaceLab_NVIDIA_internalDeepFaceLabcoreleraslayersConv2D.py", line 99, in forward

    x = tf.nn.conv2d(x, weight, self.strides, 'VALID', dilations=self.dilations, data_format=nn.data_format)

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonopsgen_nn_ops.py", line 1026, in conv2d

    data_format=data_format, dilations=dilations, name=name)

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonframeworkop_def_library.py", line 788, in _apply_op_helper

    op_def=op_def)

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonutildeprecation.py", line 507, in new_func

    return func(*args, **kwargs)

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonframeworkops.py", line 3300, in create_op

    op_def=op_def)

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonframeworkops.py", line 1823, in _init_

    control_input_ops)

    File "D:DeepFaceLab_NVIDIA_internalpython-3.6.8libsite-packagestensorflowpythonframeworkops.py", line 1662, in _create_c_op

    raise ValueError(str(e))

    ValueError: Dimensions must be equal, but are 42 and 40 for 'Conv2D_1' (op: 'Conv2D') with input shapes: [?,42,68,68], [5,5,40,84].

  16. Uhm..i downloaded the master from that link,and that looks nothing like that in the video.Am i missing something?How do you download the whole pack with 1 click there?

  17. Great job. Wish you all the best. The Gospel is 1Corinthians 15:1-4

Leave a Reply

Your email address will not be published. Required fields are marked *