If you’re an anime enthusiast but still haven’t found that special someone, you may be in luck. Computer scientists at Fudan University, Carnegie Mellon University, Tongji University, and Stony Brook University have collaborated on a tool known as MakeGirls.moe, an AI-based network able to render high-quality anime characters.
Though not the first contender in the anime generation market, MakeGirls.moe aims to create “a model that produces anime faces at high quality with [a] promising rate of success.” Aggregating pictures from games on Getchu, the site sorts features by metadata through image estimation software, allowing users to select from a multitude of facial and clothing options. These datasets are then modeled through a generator that interprets the user’s customization selections, removes unnecessary pieces of the images, and assembles the model.
The source code is available on Github, general information is available on the official blog, and additional technical information is available via the official report. Unfortunately, custom dakimakuras have yet to be announced.
Image Source: MakeGirls.moe