The success of the AI nude generator was too much for the small development team: first, the download server crashed and then the team announced the end of the project. With Deepnude GAN’s images of clothed women could be transformed into pictures of naked women.
The developers of the machine-learning app Deepnude have announced the end of the project. “The world is not ready for Deepnude,” team member Alberto writes on Twitter. The developers also seem to follow their conscience a bit: “We do not want to make any money like that.” With the $ 50 software, pictures of clothed women can be transformed so that they appear naked. The method is reminiscent of other projects like deepfakes , where people’s faces are modeled on other people. In Deepnude, however, there are body parts such as bare skin, breasts, and vagina, which are placed over the clothing.
The popularity of the software has exceeded the expectations of the developers drastically. About 500,000 people are said to have downloaded the software from the server. “If 500,000 people use it, chances are they’ll use it wrong,” say the developers. The server was also broken down under the burden of inquiries, which led to the failure of the download service. The team was not prepared for the fast viral status of the app.
Pictures of women easy to find
Deepnude uses Generative Adversarial Networks (GANs), which have been trained using a variety of images from mostly pornographic content on the Internet. The finished application works only with female bodies because according to developer Alberto it is easy to find them in quantity. The program would not be so easily transferred to the opposite sex for obvious anatomical reasons, stature, proportions and gender characteristics, unless another GAN is trained with male source images.
Owners of the application can no longer upgrade to the premium version with the announcement. Money can be demanded back on request. There will also be no new versions of the software. But that will not stop the Internet from spreading it – the developer is aware of that.