Not only cosmetics, but we are also developing towards unifying expressions in videos and from different angles in the future. However, this time, we&#39;ll try to make some changes to the clothes and expressions as well. Maintaining a consistent sense of facial features while generating is still under development and a hot topic, so there are often challenges in achieving it. <img src="https://assets.blueqat.com/public/uploads/us-east-2:4805ff4b-c3cc-4344-b165-86544c34d0bf/2024/03/25/00016-1621783070.png"/> This time, I wanted to add an expression, so I made it a smiling face, had them wear a sweater, and put on a necklace.It&#39;s become a bit difficult to tell if it&#39;s the same person. <img src="https://assets.blueqat.com/public/uploads/us-east-2:4805ff4b-c3cc-4344-b165-86544c34d0bf/2024/03/25/00040-3819500300.png"/> <img src="https://assets.blueqat.com/public/uploads/us-east-2:4805ff4b-c3cc-4344-b165-86544c34d0bf/2024/03/25/00039-4188523228.png"/> <img src="https://assets.blueqat.com/public/uploads/us-east-2:4805ff4b-c3cc-4344-b165-86544c34d0bf/2024/03/25/00043-2800078222.png"/> I feel that images of smiling faces tend to have a sense of unity. It seems important to train data specialized for portraits.I feel that we are gradually becoming able to decide on a style or add decorations starting from a bare-faced state.

Changing the style and expression of portraits using generative AI

Yuichiro Minato