Make a lip-syncing deepfake in just a few hours!

  • AUTHOR: dua
  • POSTED ON: September 11, 2020

If you thought making a deepfake was hard, then you should read this blog carefully as we’re going to tell you how to make a lip-syncing deepfake in only a couple of hours!

Usually, people use
prepackaged apps to make a deepfake for themselves, but do you know that by
following a straightforward method, you can create an instant lip-sync
deepfake? Okay, so to get started, you need to have the following: audio and
video clip of the subject.

Okay once you have the
video and the audio you want it to follow, you need to chime in the code, and
you have your deepfake ready! 

Now you can use these
deepfakes to bring down democracy or whatever – it’s totally up to you, but we will
say that this activity is extremely fun

when you have absolutely nothing to do.

James Kelleher is an
Irish designer who had
the Queen’s deepfake
. He had used the same method that was
shared online by some AI researchers. The researchers had also shared the
description of the method a couple of weeks ago in addition to the demo so that
anyone and everyone can put it on trial.

Of course, nothing good
comes for free in the world. You now have to register yourself in order to
access it. But again, K R Prajwal of IIIT Hyderabad has conveyed that this step
is to ensure no malicious usage occurs because of it. 

definitely acknowledge the concern of people being able to use these tools
freely, and thus, we strongly suggest the users of the code and website to
clearly present the videos as synthetic”
– He said.

He said that the code
is there to“encourage fruitful research
on systems that can effectively combat misuse.”

You can use the online
demo to produce a deepfake but we’re telling you nothing will happen and your hard
work will go to waste. You can download a video using QuickTime’s screen record
function along with audio utilizing
However, you won’t get any results even after plugging both the files together
into the site.

Even reducing their
resolution won’t give effective results. You will need technical grasp to
understand the problem, but if you’re here, we’re guessing you don’t have any –
and that’s totally fine!

Here’s what you need to
do: rename the files and remove any excess space present.

Now by following the
method above, you can get the work done. But if you need improved results, do
this: Use Github and access the underlying code. However, to run it, you should
utilize Google Colab which is equivalent to Google Docs but is actually used to
execute machine learning projects in the cloud. Now make sure you follow all
the steps laid out by the author otherwise your project can be subject to a series of unfortunate incidents.

Hence, don’t log in
from two different Google accounts or download the Wav2Lip Model instead of
Wav2Lip + GAN versions. Also, pay special attention to your spellings.  And if nothing works and you’re stuck, follow
YouTube tutorials that can guide you on how to solve minor mistakes.

Anyway, from this
experience, you’re going to get grip on a few things. It may be easy to get
your hands on deepfakes online, but it’s not easy to create them. The algorithms
have been laying around on the internet for years, but if you’re an amateur,
you really need to follow the traditional route of video editing. The results
will be quicker and far more convincing.


Updated September 11, 2020
Back To Top