What do you think is the most difficult part of creating deepfakes? —To realize that you need it and figure out what result you want to achieve. Let's assume you already have a great deepfake idea, and you just need to understand how to bring it to life.
Today, we'll break down how to create deepfakes with minimal effort. If you're worried that you’ll need to think a lot, you can relax. Or not! Iit depends on how serious your intentions are.
What is a Deepfake?
A deepfake is a technology that uses artificial intelligence to synthesize realistic images and videos. To create a high-quality deepfake, deep learning algorithms analyze numerous photos of a person, as well as their unique features, movements, and voice. Using this knowledge, the AI combines all the pieces to make a fake but realistic photo or video.
How to Approach Creating a Deepfake?
Since you've already thought everything through and know what you want, start with these steps:
Record a Video or Find a Suitable Clip Online: This will be your source for all your creative ideas and desires.
Create Audio with the Required Text and Voice.
Alter the Image in the Video (face, lip movements, or both — depending on your idea and chosen software).
Keep in mind that the choice of software will also affect the sequence of steps and their specifics.
If the first step is clear, you'll need to work on the following steps, so let's go into more detail.
Creating Audio for a Deepfake
For Apple devices, there is a good application called voice.ai. It is free and quite simple to use.
To begin, give the application access to the microphone and click on the icon with the desired voice. Not all personalities are recognizable from stylized images, so you will likely need more than one attempt to find the right voice from the available options.
Next, adjust the pitch and record the audio with the desired text — the recording time is limited. If you want to create an extensive deepfake, you will either need to splice the audio together or choose another program, such as ElevenLabs (though it is paid).
After processing the voice, download the finished file to your device, preferably in the gallery, for further work.
Inserting Audio into the Source Video
Now that you have the two puzzle pieces, you can combine them. Any video editor with the following capabilities will work:
- Extracting audio from video
- Inserting audio tracks
- Editing audio
For simplicity, you can use the familiar CapCut:
1. Click on Sounds.
2. Tap on the folder icon, then Extracted → Extract audio from video.
3. Select your audio masterpiece from the gallery (which is in video format) and insert the track into the source video so that the audio aligns with the lip movements.
4. Save the resulting video.
If the match isn't perfect, you can slightly adjust the speed of the audio track — either increase or decrease it. This change won't be noticeable to the ear, as long as the text in the source and the generated audio were spoken at approximately the same pace.
Replacing the Face
Now for the most interesting — giving your source a new face. For this, you can use a program like Swapface. However, you'll need to install it on Windows and register first.
After registration and all the formalities, you have a few free attempts to generate something amazing:
Select video Faceswap mode.
Upload your source video into the program.
Choose an existing model (mask) or upload your own. When uploading, it's important that the program recognizes the face of the model; otherwise, it won't be usable.
Start the generation process and wait a few minutes.
You did it!
If the result didn't impress you much, don't get discouraged — it's all a matter of practice: improve your skills, discover new tricks. Let's find out what experienced webmasters think about deepfakes.
ZM Team’s Expert Opinion
We asked the ZM team a few questions about their attitude towards using deepfakes for ad campaigns.
‘Skillful use of an influencer's face that fits the context is definitely an advantage in terms of trust’ — ZM team
Does your team create deepfakes for creatives or is it in your plans?
ZM: Currently, deepfakes aka celebrities are a popular approach in gambling creatives, which can be assembled in various ways: full-fledged deepfakes, just lip-syncing, or skillful editing of real materials with the desired influencer. However, our teams do not use this approach; we prefer other methods.
Do you consider using deepfakes in ad campaigns promising and why?
ZM: Yes, this is definitely an important direction, including in white marketing, where deepfakes or licensed characters are used, or their own are created, allowing for quick and convenient assembly of videos with almost any content. Therefore, many will definitely use this; it is in demand.
What are the advantages of advertising with deepfakes?
ZM: The technology itself is highly promising, especially for digital marketing. Many companies have already simplified content creation with avatars. Moreover, the technologies are developing very rapidly: the quality of deepfakes is improving month by month, and their production is being reduced to a few clicks.
In iGaming, skillful use of an influencer's face that fits the context is a clear plus for trust, and hence for conversion.
Do you think there could be problems with platforms and placing deepfake creatives on them? Is there a way to avoid these problems without sacrificing great creatives?
ZM: Many platforms, including our most beloved Facebook, are already planning to actively regulate AI content: they are beginning to require special tags, separate categories, and much more. This also applies to advertising, but since these processes are not yet fully developed, it is hard to say what this will lead to in the end. Therefore, at the moment, it is still possible to place deepfake creatives, but in the future, there will likely be special tags and audience restrictions for AI-created "people."
What do you think are other disadvantages of deepfakes?
ZM: Using deepfakes of famous personalities in the CPA sphere is highly controversial, which is why ZM media buying teams do not work with this approach in their creatives.
The reasons are obvious: not only the possible problems with platforms and advertiser bans that do not approve such methods, but also the legal responsibility for violations, which can be quite real, contrary to many webmasters' beliefs in their absolute anonymity and safety.
To sum up…
As you can see, deepfakes are indeed a promising topic and, apparently, a whole niche in advertising with its own rules and regulations in the near future. So, if you haven't started mastering deepfake technologies yet, now is the time to do so to avoid running after the departing train later.
Share the article
No comments yet. Be the first!