Face swapping with ai
Personal Tests
Another interesting thing we can do with AI is swapping faces on generations. I ran extensive tests on this tool on my own face.
The process starts by gathering a comprehensive enough dataset of a person’s face. In my case, I gathered selfies, and some professionally shot portraits, from as many angles as possible so that the AI would “get to know me”, and remember and understand my face. I used 50 photos for the data set. It works with fewer but more images is more info for the AI in this case. I normally wear lipstick so this made so many generations come out with makeup on, this was an interesting find.
In the end, I was able to produce“photoshoots” by world-renowned photographers and artists from the past, in dream-like locations, wearing couture or fantasy wardrobe. Anything I asked for.
It’s a fun tool and I can see many exciting uses for it in the future.
Images were created on Stable Diffusion + background extension, retouching, and extra compositing/cleanup were done manually in Photoshop.
Some of the images of my face I provided Stable Diffusion to learn and get to know my face. This is ME
Generations with face swapping. This is not (?) ME… technically
CASE STUDY FOR VICCI EYEWEAR
Vicci approached me to see if we could expand their portfolio of images of their product shot on models but generating the women with AI. I was tasked with generating elegant studio photos, in which the product can be seen in detail, for a target audience of women aged 35-50.
I worked generating my own pose references with photos from my portfolio and images provided by the client. To base my generations on images that the client already has uploaded to their website I manipulated them first to change the model’s factions and hair so that I could achieve truly different looking women.
