Elan Mitsua is an AI VTuber
with the theme of “Art made
with Everyone”.
Mitsua Likes, an image
generation AI
trained on contributors’ “likes”,
will shape her aesthetic sense
to create unique art.
with the theme of "Art made
with AI and everyone" AI VTuber majoring in Art
-
Exploring Expressions Unique to AI
An artistic expression grounded on two powerful characteristics: the characteristic of a collection of likes, derived from training on images provided by participants, and the characteristic of artistic expression unique to AI, which differs from human expression.
-
Everyone Can Become an "Art Teacher"
Mitsua-chan grows by learning from 'art teachers'. Anyone can become an art teacher . To provide a secure and comfortable environment for the teachers, we are actively addressing ethical issues surrounding generative AI.
- Discord server
- Mitsua Contributors
- "Mitsua Contributors" is a place where art teachers engage in various topics, ranging from training image submission and image generation to everyday chats and serious discussions with community members. Even if you haven't joined yet, you can view training images and post comments in public channels simply by agreeing to the terms. Please feel free to join this server if you are interested in Mitsua-chan’s activities.
Opt-in Image Generation AI Mitsua Likes
-
Trained only on Rights-Cleared Data
We have developed a completely original model without using any existing foundation models. We collect and train on data that has been explicitly authorized through an opt-in process, as well as verified public domain and open-license data, in accordance with contracts and regulations.
-
No Training on AI-Generated Data
The training data does not include images generated by other image generation AIs or texts generated by LLMs/VLMs.
-
All Training Data is Open
All training data is made public on this website (partially) and on our Discord server, ensuring transparency.
-
Attribution for Generated Data
Credit display is mandatory for generated data to identify the original art teachers from the training data and prevent issues such as falsely claiming of hand-drawn works.
"Opt-in" means training on data that has been provided with explicit consent to be used for training. Additionally, “Mitsua Likes” trains on these authorized images and works as well as materials with open licenses, public domain, CC0, and other works where the creators have waived their rights.
CERTIFICATION
Mitsua Likes is certified as an AI
model
that is not
trained on
copyrighted works
without a license.
The Mitsua Likes model, an entire architecture of a text-to-image generative model, is Fairly Trained certified, but all our previous models including Mitsua Diffusion CC0 and Mitsua Diffusion One are not covered by Fairly Trained certification.
This indicates that a third-party organization has certified the Mitsua Likes model does not perform any unauthorized training processes and does respect the rights of content creators.
The Mitsua Likes is the first generative AI model developed by a Japanese company to have been certified by Fairly Trained.
News
-
2025.6.30
"Mitsua Likes" 返礼品デジタルコースのお知らせ
-
2025.3.21
「絵藍ミツア」及び弊社に対する誹謗中傷等への対応について
-
2024.12.17
基盤モデル不使用のオプトイン画像生成AI「Mitsua Likes」公開!
-
2024.4.23
【受付終了】 “Mitsua Likes” 返礼品のおしらせ
-
2024.2.19
“Mitsua Likes” 学習データ受付延長とポーズ&3D小物受付開始!
-
2023.11.27
“Mitsua Likes”- 絵藍ミツア 新モデル学習プロジェクト始動!
-
2023.10.17
【受付終了】 返礼品グッズのおしらせ
-
2023.8.17
OpenCLIP使用のご報告と今後の方針
-
2023.4.5
学習モデル公開の理由と悪用リスクの考え方
-
2023.3.1
素材を利用した学習画像投稿への注意喚起