Abstract Engine Co., Ltd. has released Japan’s first fully trained and certified AI image generation model. The service, called “Mitsua Likes,” is headed by the company’s AI VTuber Eran Mitsua. Based on the theme of “art made by everyone,” the concept is that users ethically provide art school student Eran Mitsua with images to learn from, which in turn influences what the AI model creates. Officially certified by the US non-profit organization Fairly Trained (as reported by Otaku Research Institute).
Mitoa Likes is not based on existing AI models and is trained only on images that have been explicitly approved or verified to be in the public domain. AI-generated data from other AI learning models is also not trained. To ensure transparency, all training data will be published on the official Discord. Additionally, images produced by the model are clearly credited to identify the creator of the source image.
To date, Mitsua Likes has been trained on over 58,000 images. Starting in January 2025, the company will begin accepting submissions of new images for AI to learn from. For more information about “Mitsua Likes,” please visit the official English website.
