Deepfakes Are Changing The Way We Meme

If you told someone 20 years ago that you could make it look like the president was reading an offensive joke with near 100 percent accuracy, you’d be laughed out of the room. Now, that very technology has reached consumer-grade levels, allowing anyone with a modicum computer literacy and time to create what’s known as a deepfake.

The term, a portmanteau of "deep learning" and "fake," refers to deceptively realistic videos created with digital imaging technology powered by machine learning algorithms and artificial intelligence. In simple terms, this involves feeding a bot a trove of videos, photos and audio clips in order to map a convincing digital likeness of a subject over another person's face.

While deepfake technology has mostly been used for films at a professional level, and sexually explicit content at the consumer level, memers are now starting to realize the full potential of such software and apps for comedic purpose. We’re currently witnessing the rise of deepfake memes, and this is just the beginning.

The History of Deepfake Memes

Fakery has always been a part of meme culture. Faking tweets and social media posts is a common avenue for memers, and one they perfected quickly. It used to be that you needed to learn Photoshop to do this successfully, but now countless, simpler computer programs and apps can achieve the same results. Deepfaking looks to take this to the next level by allowing memers to create videos so authentically fake – from voice to appearance – that nobody knows whether they’re watching something real or not.

Baka Mitai facemorphs have been an explosive form of synthetic media bearing many similarities to deepfakes. The memes use facemorphing applications to make any subject appear to sing along to the song "Baka Mitai," popularized in Yakuza 0. The first one made it appear as though YandereDev was singing to the song, and from there, the roster has only expanded.

The most popular variant sees nine different meme targets in split-screen, including Jeffery Epstein, YandereDev, the Staring Hamster and Craig the Brute, singing along to the song in perfect synchronization. It’s not just a funny meme, it’s also absurd and technically fascinating.

Deepfake-style technology has deep roots in other forms of synthetic media, including speech synthesis and face swaps. With voice-faking, we’ve seen users make Barack Obama read the Navy Seal Copypasta, and Joe Rogan espouse the benefits of an all-ape basketball team, to name a few. The swaps are eerily realistic, with only slight tonal changes breaking the illusion. If someone with no knowledge of deepfakes listened to these, they might think they’re real audio clips, despite how absurd the content is.

Current examples like these give us a peek at the fast-approaching future of deepfake memes. YouTuber French Faker, for example, exclusively makes deepfake memes. Some highlights from his channel include 6is9ine as Filthy Frank playing the nose flute, and a faux presidential advertisement for Kanye West featuring the rapper’s face on a Ricardo Milos dance video. French Faker isn't the only account doing this, and it proves that you don’t need some massive following or position to create convincing deepfake memes. Channels like derpfakes and Ctrl Shift Face are some of the more prominent creators of this content, but a quick search on YouTube will result in many others, alluding to the rise in such meme-oriented deepfakes.

Deepfaking Goes Consumer

The most popular deepfake-style technology is Faceswap, a program that lets you swap the faces of anyone. There’s an in-depth article describing how it works over at ArsTechnica, so if you want all the details from someone who's qualified, you can check that out. The writer, without any prior deepfake experience, is able to make an example replacing Mark Zuckerberg's face with Data’s from Star Trek. The mouth matches up, the eyes move around – it’s eerie how well the whole thing comes out. It’s not perfect, but it’s impressive, especially from someone with no prior experience.

However, Faceswap still takes a certain amount of technical know-how and a good computer to use. At a simpler level, users can access sites like fifteen.ai – currently down, but soon to be revived – to deepfake voices. Examples on the site include SpongeBob reading the New Anime Plot copypasta, and GLaDOS reading the Rick and Morty Copypasta. By simply typing what you want, you can make it sound like certain characters are saying anything in mere minutes.

Doublicat, also a face-swap app, is another way that deepfaking has reached the consumer level. It’s available on iOS and Android and lets the user create impressive deepfake videos with nothing but a source photo and video. While the quality isn’t professional-grade, it’s still impressive tech that we couldn’t have dreamed of only a decade ago, and it’s right in everyone’s pocket. Besides, deepfake memes don’t have to be perfect – they may even be funnier because of their imperfections. Deepfake memes only need to be “meme-quality,” and that level has already been achieved.

Just look at the popularity of the deep-fried, low-res memes that litter Instagram's Discover page. In many cases, subpar quality is a part of their charm and success. Early consumer quality deepfakes could set the precedent for deepfake memes going forward. When people start trying too hard, especially with memes, others tend to lose interest. Considering this, deepfake tech is kind of perfect for memes as it currently stands. It’s when it starts to get better – and when people stop using it for comedy – that things might get weird. Thankfully, meme humor is pretty easy to spot. We probably won’t have too much trouble telling the jokes apart from the serious attempts at trickery.

The (Potentially) Scary Future of Deepfakes

Many people are afraid of the possibilities of deepfake technology. They imagine a future where anyone can make important figures appear to have said or done something they never did, and share that as the truth. As of now, consumer-grade deepfake technology isn’t likely to pull this off. The Uncanny Valley effect is still very much there visually and audio-wise, no matter how good the results. For this reason, most deepfake memes will likely be pretty obvious for quite some time. Most memers don’t have access to the most realistic deepfake tech, and even if they did, they’d still need the know-how. The chances they’ll waste that knowledge creating 100 percent realistic deepfake memes seems unlikely.

Using deepfake technology to meme is inevitable, but so is using it for more nefarious deeds. That’s why it’s important that we get to know these types of memes and to understand what a deepfake sounds and looks like. If deepfake “crimes” and trickery really do take off, memes using the tech might be a great way to introduce people to deepfakes and how they look and act. At the very least we're looking at a whole new avenue for memers to explore, and we could see some truly great original memes come out of the deepfake revolution.

Meme Insider is a Know Your Meme publication and the world's leading internet trends and culture magazine. Find out how to get your first print copy for free, and check out the Meme Insider website for more info.