TL; DR

Not all applications claiming to be free truly offer their services without cost. Among the top applications listed in Google search results for the keywords 'free deepfake maker', Genice and Vidnoz stand out as genuinely free options. However, it's important to note that Vidnoz has limitations, allowing only 8 image attempts and 3 minutes of video usage per day. In contrast, Genice offers unlimited usage for both images and videos as of 2024. While the quality between them is not significantly different, there are certain characteristics that distinguish them in terms of results. Therefore, it's advisable to try both applications and choose the one that best suits your preferences.

Introduction

What Are Deepfakes?

"Deepfake" can be used both as a verb and a noun. When used as a noun, it refers to images or videos that are not real and are intentionally created to appear authentic. They are often indistinguishable by humans. When used as a verb, it describes the action or behavior of creating these fake media. The term "deep" in "deepfake" signifies its incorporation of deep learning technology, where deeply trained neural networks produce more convincing results than previous methods. However, there are various technological approaches to creating deepfakes, even within deep learning methods. As a result, different services offering deepfakes exhibit varying quality and characteristics.

Why Did I Start This?

The Google search engine is generally reliable, but occasionally it displays pages in its search results that are not entirely accurate. This is because Google cannot verify all the information presented on every page. Therefore, I don't blame Google for this limitation, and it sparked my curiosity about services that offer free deepfake creation. Not everyone has the luxury of easily spending money, so I decided to start my journey by testing free services. While our intuition may suggest that free services may not be as good as commercial ones, I wanted to verify if they truly lack in quality.

What I Will Do for This

Although I still believe that Google prioritizes better services at the top of its search results, I decided to visit three services listed under the search term "free deepfake maker." As shown in the screenshot below, we have three websites: Hoodem, Vidnoz, and Deepfakes Web. I will explore each site to assess their functionality and the quality of their results. If I am unsatisfied with their performance, I may question Google's ranking and consider that the pages listed below the top ranks might offer better alternatives. However, I will not cover these alternatives in this article; instead, I will save them for future investigations.

Services

Hoodem

Hoodem is at the top of the list, so I was really excited to explore it first. However, when I visited the website, a loading icon appeared and I had to wait quite a while, which was unexpected given the fast response times of most websites these days.

Apart from this, as soon as I landed on the page, a demo appeared. It allowed me to select sample faces on the left and a target video on the right, with the selected face being replaced by the one I chose. It was quite interesting.

I chose a face similar to Emma's and a video featuring Donald. The result? You can find it below. Honestly, if you hadn't mentioned that it was swapped with Emma, I wouldn't have been able to tell.

I suspected that the quality issue occurred because it's a male face being replaced with a female one. So I changed the video to another one featuring a woman. You can see the result below. Once again, I wouldn't be able to tell who this woman is without any hints.

Then maybe this is only an issue with Emma. I chose Elsa because I really like her. Take a look at the result. While some aspects of the eyebrows resemble Elsa's, the mouth shape seems to be different.

I decided to give it a try myself, so I signed in and uploaded a video to create a deepfake. The process seemed straightforward: upload a video, add a face to it, and then download the result. Sounds good.

When I reached the video upload page, I realized that I couldn't proceed further because it showed that I had 0 credits and prompted me to the payment page. I attempted to find coupons or trial offers but couldn't find any, so I decided to stop testing this service further. I'm surprised that even the first service couldn't satisfy me at all.

Vidnoz

Vidnoz appeared to be usable free of charge. They offered options for deepfake in both images and videos, so I opted for the image-based service, assuming that image-based swapping would be the basis of their video generation. I used two template images they provided and uploaded my target faces, starting with Emma. You can see the original images and the swapped results below. The results seemed quite good, as they were easily recognizable without any hints.

‍

‍

Next, I tried swapping a face with Jennifer's. The result still looked good enough to be identified as her. However, after a few tries, I noticed that the service wasn't entirely free. I would need to pay if I wanted to create more than 8 deepfakes per day. Still, as a hobbyist, it wasn't too bad.

‍

The service only required a single image of the target face, making it very easy to use. However, I also felt a little worried because it means that sometimes we may not be able to provide enough details of the face we want to deepfake. I will revisit them in the future to assess how accurately they represent the target person in the result with only a single input.

Deepfakes web

After encountering the first service, which wasn't free at all, I decided to see if I could use this service for free before exploring any examples they provided. Upon signing up and accessing the page for creating a deepfake, I was prompted to upload source and target images or videos. However, when I attempted to upload my source image, I was asked to purchase GPU credits, indicating that even a trial use required payment. Once again, I was surprised by the lack of free options. Upon revisiting the search result description, I realized that they never claimed to offer free options; it seems Google had selected them somehow.

‍

‍

Genice

Genice is a free application as of 2024 for both image and video generation. I decided to give it a try using the same target image and source face that I used in Vidnoz, which was the only available option for us. You can see the deepfake result below. The quality of the output from Genice is not any worse than the result from Vidnoz at all. Honestly, I couldn’t tell which one is better, so I will let you decide for yourself.

One difference is the way they generate results. Genice accepts multiple source faces, using multiple images to understand which face to create. I gave them 5 images of the source face for each try, but you can also use a single image if you prefer. Another difference is that Vidnoz refused to accept Taylor’s face at all, suggesting there may be certain guidelines they adhere to. In contrast, Genice accepts any images you want. While I couldn’t compare the results for Taylor here, I can share the output from Genice below.

Additionally, Genice required training a model to swap the face in the target image. This process took almost a few minutes to complete, but once it was done, generating an image only took a few seconds.

‍

‍

Comparing Quality

When comparing the characteristics of the outputs from Vidnoz and Genice, the results from Vidnoz seem to retain the original facial expressions more than Genice, whereas the results from Genice appear to be more realistic to the person. For instance, we've never seen the source face smile in that particular way before. It seems that Vidnoz generates more virtual faces that don’t exist in the real world. However, there may be someone who prefers this style, so I wouldn’t say which one is always better or satisfies your needs. Overall, they both generally look natural, and no one would easily say they were generated by AI.

Why Aren't There Many Free Services?

One might wonder why there aren't many services providing this for free, especially when you consider that many other non-AI-based services are offered for free. The primary reason is that these services require GPUs to process user requests. As you may know, the price of GPUs skyrocketed a few years ago due to cryptocurrency. This surge in demand made it challenging for service providers to afford GPUs. Sometimes, they couldn't even buy or rent them from cloud services due to high demand and limited supply.

There was a time when Google Colab was used for this purpose, providing GPUs to users for free. Many people flocked to it for deepfakes. However, Google now officially blocks any usage of Colab for deepfakes, rendering it invalid. Given this situation, services for deepfakes without any cost are very precious to us. While we don't know how they sustain their costs, there may be some ways they manage to do so.

If you understand this background, you may also consider trying services that require payment, as there's no such thing as a free lunch, and they need money to prepare your lunch. Another option for accessing free deepfakes is by utilizing open-source software on your computer with a GPU. However, you'll need an initial investment to purchase them.

Paid Deepfake Services

There are numerous paid services available for creating deepfakes. A simple Google search using the keyword "deepfake app" yields around 20 apps, including Lensa AI, Reface, MyHeritage, Deep Art, and many others. Some of these apps are even available as smartphone applications. While I can't guarantee their quality since they require payment, I hope they provide reliable and higher-quality results. I plan to review them in future work.

I believe that if these services are truly confident in the quality of their results, they should offer trial versions for users to test. Therefore, don't hesitate to look for services offering trials to assess sample results before making any payments.

Is it Safe to Login with an Email Address?

Yes, it's safe. Your email address isn't displayed publicly anywhere on the service, unlike community forums. It's simply required to notify you once your request is completed. Since tasks involving GPUs can take anywhere from a few minutes to hours, you're likely to navigate away from the site before the results are ready. Therefore, receiving an email when the process is complete is a convenient way to stay informed. You don't need to harbor suspicions about providing your email address; it's genuinely necessary for the service provider.

Conclusion

Our exploration began with a search for "free deepfake maker," revealing that Google's rankings can sometimes mislead users into thinking certain services are free when they're not. To save time, it's advisable to opt for genuinely free services like Genice or Vidnoz. Despite Google's ranking, these services offer authentic free usage.

Among these free services, we noticed distinct differences in the results they produce. While we won't designate one as superior, personal preferences play a significant role.

Given the high cost and limited availability of GPUs, many service providers struggle to offer free services. Therefore, make the most of these free services while they're available. If you're serious about generating deepfakes, paid services may offer convenience and higher quality.

When creating deepfakes, exercise caution and refrain from sharing them online without obtaining consent from individuals involved. If you intend to use them professionally, always seek permission from the relevant individuals.

Deepfakes represent the fusion of imagination and the visible world, akin to moving from reading a book to watching a movie. While advancements in technology may spark concerns, history shows that we've always managed to harness them for the betterment of humankind. So, don't be afraid to explore; give it a try.