I’ve previously used AI to clone my voice, so why not try to clone my appearance, voice and everything you normally see on video for that matter? In other words, I set out to create a deepfake.
Let’s take a look at the process and when a deepfake might come in handy. I discuss:
- What’s a deepfake?
- The point of creating one
- Are they realistic?
- AI platforms to use to create a deepfake
- How to create your deepfake
- Ethical considerations
What’s a deepfake?
Regardless of the use case, a deepfake uses artificial intelligence technology to create a realistic-looking version of a person that includes some kind of spoken words – typically ones that the person hasn’t actually spoken. In this case, a deepfake of football coach Brian Ferentz says something Brian never said about doubters.
Another scenario could bewhere I create a deepfake of myself. Here’s an example of my deepfake reading a section from my thought leadership article. I created this with Hey Gen.
What’s the point of creating a deepfake of yourself?
If you are a video podcaster or create a lot of other video content based on written scripts, using a deepfake can actually help you churn out way more of that content. As long as you have that written source content, you can upload it into Hey Gen, which will then create the AI clone video version of you.
Are the AI versions realistic?
I thought my version was pretty realistic and close. My youngest daughter didn’t think so. Would the audience notice or even care if it’s not perfect? It probably depends on the value of the content as well. A great delivery will not save crap content, so the content needs to be useful and relevant, and the delivery needs to be close, if not better.
Read next: 8 Tips For Using AI to Write a Book
AI platforms to use to create a deepfake
The leader in this field has been Sythnesia, which I first used in January 2023 to create a live podcast.
A Synthethia alternative is Hey Gen. Here’s a quick overview of differences by plan:
|Lowest Paid Plan Price
|$24/month billed yearly (Creator)
|$22/month billed yearly (Starter)
|Highest Paid Plan
|Enterprise (Contact for Pricing)
|Enterprise (Contact for Pricing)
|Included Seats in Lowest Paid Plan
|1 editor + 3 guests
|Included Video Duration in Lowest Paid Plan
|120 minutes per year
|Custom Avatars in Lowest Paid Plan
|Unlimited photo avatars
|1 webcam avatar (more are paid add-on)
|Languages and Voices
|Yes in higher plans
|Stock Media Library
|Yes in higher plans
|Live Chat, Email
In summary, Hey Gen has a free tier while Synthesia does not, but Synthesia’s lowest-paid plan includes more features like a screen recorder, captions, and a stock media library. However, Hey Gen offers more customization of avatars. Both have enterprise pricing for large teams and budgets.
How to create your deepfake
Let me show you in Hey Gen, but know it’s very similar in Synthesia. The process in Hey Gen is quite simple. You can even give it a test run before getting a paid subscription. Login after creating your account.
From there, record a sample clip. Best practice tips:
- Record in your normal environment as that will be the backdrop. I simply recorded in my usual studio and office spot.
- Be yourself. Look at the camera at times, use hand gestures – if that’s your thing; and talk like you normally would.
- To be natural, pick a topic and talk about that for the prescribed time frame (30 seconds to 2 minutes depending on sample). That ensures it’s not too scripted and really brings your usual self out. The content doesn’t have to be perfect as it’s just a sample, but I just tried to talk about the topic for the required time period.
Then create your first video. Simply upload the transcript or type out what you want your deepfake to say.
Once ready, submit the project and the video is created.
From there, you can use it where you need to. Super easy process – as long as you have the source content.
Hey Gen us also integrated into Canva – under apps.
It should go unsaid but clearly you should not create deepfakes of other people, but using yourself to level up content creation – what’s wrong with that? Nothing in my opinion.
It might get a bit trickier if you are using a deepfake of yourself on a company’s behalf. I would say that common sense would dictate that they can keep using content that was created while working for them. After all, blog content written for a company while working for them doesn’t follow employees, usually podcast recordings also don’t. So no different here in theory. What’s been done has been done, but if they want to create new content with your deepfake some kind of agreement or licensing deal should be drawn up.
To be really sure, ensure that this is covered in terms – usually it is as most company’s have a line with content teams that says something like: “Anything created for the company while working for the company was created on the company’s behalf and belongs to the company.”
Read next: Is marketing a good career?
Like any new technology, it’s usually good to explore and that’s exactly what I did here. Will I use this going forward? Not sure, I don’t see myself loading up my deepfake to do podcast interviews with a live guest, but I suppose you could create a video interview podcast where one deepfake interviews another.
And the use case of creating more video content based on existing written scripts is definitely an option.