Making an AI-generated selfie was a challenging but fascinating process for me. The hardest part was definitely trying to describe my own face. After thinking about it for a long time, I finally went with this prompt: “ A young Chinese woman wearing brown-framed glasses, her black, shoulder-length hair styled without bangs. She has a soft, rounded face with a narrow forehead and a gentle, fleshy nose. The image blends these natural features with a high-tech, cyberpunk aesthetic.” I wasn’t sure if the AI would really “get” me, but it felt like the best way to put it into words.
I tested this prompt across three platforms. NovelAI pulled me into Japanese anime territory, making me look like a manga character. It was cute, but it completely ignored my “no bangs” instruction and “cyberpunk aesthetic.” It also felt stereotypical because the AI seems to think that anyone with yellow skin wearing a cheongsam is Chinese. By doing this, I got a clear look at different racial stereotypes regarding Chinese identity. I get why it happens. It’s usually hard to tell East Asians apart just by facial features, so the AI relies on these external “decorations” to signal race. I also tried Gentube. It did the same thing as NovelAI. Finally, I tried Gemini. My first attempt wasn’t great, but at least it finally got rid of the bangs. The second one looked better, but it still lacked my personal spark. I think it would be much faster to just upload a real selfie, but I’m worried about privacy. I don’t really want my face to be sucked into an algorithm and used to help create more stereotypical AI images of Chinese women.
Figure 1. generated by NovelAI
Figure 2. generated by Gentube
Figure 3. My first trial with Gemini
Figure 4. My second trial with Gemini
Using an AI to generate a selfie feels a lot like sitting in front of a slot machine. Sometimes the result surprises you. Sometimes it disappoints you. But either way, you find yourself wanting to try again. You never quite know what the next generation will bring. Maybe this one finally captures the curve of your glasses, but loses the shape of your face. The perfect result always seems just one attempt away, close enough to keep you reaching for the generate button one more time.
However, an AI generates an image of you, and what you keep encountering in the process is a self that no algorithm can fully grasp. Every gap between the output and reality is a quiet reminder that you are more complex than any string of words can capture. The real you cannot be reduced to a prompt.
A few days later, I started thinking about the prompt I used to generate that AI selfie. It was so vague which shows that back then, I did not have a clear idea of my past or my future. That is probably why my online presence feels so blurry too. I’ve realized that the first step to making a selfie is figuring out what you actually want to share or how you want others to see you. Only after you figure that out can it be turned into a real prompt.
Translating my critical analysis into a blog post is its own act of identity negotiation. For the blog, I have to strip away the jargon and citations, using a voice that feels readable, personal, and immediate. Writing for a blog also forced me to put my personal experience front and in a way academic writing resists. Because blog readers come from so many different backgrounds, you have to make sure your ideas are approachable for everyone. A good blog also needs to be scannable. To achieve this, I bolded some key sentences and tried to make my writing short, simple, and less formal.
Comments
Post a Comment