Models that do not exist – artificial intelligence in erotica.

Models that do not exist – artificial intelligence in erotica.

Artificial intelligence is increasingly entering our lives and we do not even realise how often we use it: search engines, Chatbots, speech translators, speech synthesis, virtual assistants all this is possible only because of artificial intelligence. Of course, it also came to the world of erotica created artificial intelligence-driven sex dolls, erotic gadgets, etc. It is also used in photos/videos in various ways, but sometimes in infamous ones. An example is Deepfake technology, which allows you to replace faces in a video. Swapping the faces of people, e.g. in porn materials without their consent is uncool and should be condemned, but we will find a lot of such materials on the Internet. Deepfake, however, also has a bright side, this technology has found application in feature films and commercials. I also decided to try using machine learning in graphics.

Inspired by an article on the technology blog gadzety360.pl on the generation of photorealistic graphics using artificial intelligence and the StyleGan2 algorithm, I decided to experiment in this field. However, generating only faces or pets somehow did not encourage me to experiment and devote time, but if so hmm … generate some erotica?… I decided to try and share with you the effect of this experiment. Having fairly decent computer hardware with an RTX card, I prepared it for this task.

For experiments, I chose the StyleGan2 variant in the Ada version, which does not need as many as tens of thousands of photos to achieve a decent effect. In StyleGan2 Ada, “only” a few thousand of them are enough for the computer to learn how to generate photos of decent quality and that’s all I was able to prepare. Setting up and running the algorithm wasn’t as easy as I expected, but I got through the process and started teaching the algorithm in generating erotic photos.

The training required patience and lasted many days, effectively eliminating the computer for other jobs. However, the effect I obtained, and after conducting only half of the recommended length of training, turned out to be extremely positive. I created quite realistic models of erotic toples that do not exist! At some point, looking through the effect of training in StyleGan2 Ada and the photos on which I trained, I got lost myself and I was not sure what was created by the algorithm and what is the original photograph. Below are some examples of non-existent models.

We are on an erotic blog, not a technological one, so I will not present how to conduct such a training step by step, but those interested in this topic will easily find materials on the Internet. What do you think about the women generated using artificial intelligence? I look forward to your comments.