Hello. I don't know anything about cosmetics, but I think I need to use generative AI somehow and am gradually making progress.
For cosmetics using images, I believe you would normally try by overlaying layers on top of a standard image. However, this time I've tried creating an image from scratch.
By doing so, I can accommodate various scenes and create natural-looking images.
This is an image created by a standard image-generating AI. The sampling values are kept modest.
I will apply eyeliner and lipstick to this.
I included specific makeup texts in the prompt.
Fundamentally, generative AI creates images from scratch. If the original model's data does not contain anything similar to the person, it becomes challenging to produce a similar model even if one is chosen.
- Outputting a portrait using a generative AI model
- Applying makeup to that portrait using the same generative AI model
In this case, since the image is generated from scratch every time, it works well. On the other hand,
- Inputting one's own image
- Applying makeup using a generative AI model
In this scenario, there's a high chance of getting a model that doesn't resemble the input at all.
Since the above image is also created from scratch, the clothing may end up being different. While it's possible to aim for consistency, such technology will depend on future development.
To output one's image in high quality, it's necessary to advance learning in the model. This can be achieved through fine-tuning or additional learning methods called LoRA, and I hope to expand its use according to demand.