Written in 2022! updates at the end, read on for visual journey of me getting gendered by picsart genAI lol xoxo
non-binary slow-burn cyber-body-horror with a neutral ending 💁🏻
In this blogpost I will share my experience using the AI image generative features in photo editing app PicsArt.
Disclaimer: not affiliated with or promoting PicsArt, just nerdy about it and a niche fan xoxo
tiny intro:
😽 I'm a digital artist and researcher and have done projects involving picsart, digital rights, art and technology
🔎 some links for picsart and ai context
some context:
The backdrop for all this is that I'm a recent grad of a web development bootcamp (2022), and I've studied cyber security, digital forensics and data privacy (2020).
For the past year I've been looking for my first job in tech and navigating my expression as a non-binary person through this corporate north-american lens.
In the meantime I've been creating and continuing different projects as an interdisciplinary artist and researcher, focusing on digital rights, data privacy and personal archival. I've made zines, hosted an interactive workshop, built some websites, and made a few visual art series of digital works.
I've used PicsArt for 8 years, since 2015. What started as fun way to make collages for Instagram turned into the main tool of my art practice, something personal and special. Originally I painted with gouache, and made sculptures from found materials, including cardboard. When I realized I wouldn't have a reliable space to paint for a while I decided to learn more about making digital art. The Adobe Suite intimidated me, I didn't know much about using software in general, not just as a means to create visual works.
PicsArt was my solution to this, I played around with it for a while, and let myself learn the ins and outs of just this app to see how I feel about the work I produce. That was in 2018, and this process overlapped with my Tech studies - I learned about cyber security, privacy, and developed my research on digital rights. All of these themes reflected in my artwork and formed a visual diary of my experience.
Eventually this diary component became the most important part of my relationship with technology. I feel that it's the element that keeps my voice alive as I learn and assimilate to new tools and ways of thinking, and especially as I navigate the corporate side of it all, looking for work, building portfolio projects and trying to find a place for myself in this industry.
This lead me to centering these precious elements of personal archival and expression in my research on digital privacy.
I want to resist hyper-vigilance, isolation and self-censorship to persist as a femme.online, and figure out what that means to me in this contemporary moment.
the setup:
This project was something that came together as a mix of all these themes and experiences, started as a low-stakes PicsArt study in expression and perception - I wanted to try making a Professional Image of myself to be perceived in my situation by simultaneously : [ potential employers, industry peers, role models, non-industry peers, artists, and myself ]. I was both playing around, but also hoping to make something I can use so I don't have to think about this so much anymore.
professional image of myself:
I was working on my personal website, and wanted to have a nice pic to use of me for the "About" page that was different from the one I currently have as my profile picture on linked-in. I also made that(3) using PicsArt but with no AI features.
This was made very hastily and finished in Lightroom even though it started as a very poorly lit photo in my bedroom(1), I used the background remove and add features, the "blue"-"fabric" background blurred - in this case the shirt is mine and I was wearing it in the photo, this time I wanted to refresh my outfit professional-style, but without investing in new business casual.
This blue image(3) was also only meant to pass a busy professional glance on LinkedIn, where you can't zoom in to someone's profile photo.
This time I wanted something I could use more widely, but also more fun and reflective of me.
(loose objectives, I just wanted something low stakes to do)
change of outfit, 1st ai prompt "work shirt":
Initially I found an old pic (not shown) where I was posed in a way I could use for this project, and my hair in it is similar to how it is now. My shirt was pretty crumpled though, so I thought I would try and replace it with a "sticker" of a shirt from PicsArt's extensive "sticker library".
 |
| 4 - example of sticker library with "penguin" search |
I also wanted to replace the background because there were some decor choices that didn't resonate with me for this purpose.
I made the background purple, and gave myself a pink shadow to be cute. Then I went to get a shirt and I saw the "ai replace feature" and thought I would finally check it out.
I typed in the prompt "work shirt", and that's what I got, even with a little white circle for a company logo, so I added my own to the image. The buttons look pretty AI-generated and I appreciated that(5).
 |
| 5 |
office shirt confusion:
However, my vacant gaze from 2020 didn't sit right with me for this purpose(5), so I went back to the origins of this whole thing and used the photo I took in my bedroom for the blue-linkedin-photo project(6).
 |
| 6 |
I started by selecting the shirt area of the photo(6) and the
"ai-replace" feature, and submitting the prompt "office shirt"(7, 8).
 |
| 7 |
 |
| 8 |
 |
| 9 The
resulting image - although I love the relaxed neckline, also gave me
some chest volume that I don't have #topsurgery xoxo. | |
I'm wondering if in order to apply a shirt it needs to pick a gender to decide what silhouette + garment to generate. I wonder of this applies to clothes, "shirt",
or everything? And how strict is this, in regards to more ambiguous images?
I thought I would try the prompt again on the original photo to see what happens.
 |
| 10 (a, b) It definitely did it again but I thought I would choose the first image (10a) and work with that, I also love the skin collar. |
 |
| 11 |
business office + Hair Boss:
I then selected the "remove background" feature and used "AI background" with the prompt "Business Office" and the results were amazing (12 a, b).
 |
12 (a, b)
|
I chose the second option because I liked the hairstyle it added on for me (12b).
 |
| 13 |
The background was a little harsh (13), so I went to generate myself another business office (14a,b).
 |
14 (a,b)
|
Obsessed with the hair in the first pic (14a), but I went with the second option for the wood finish(14b).
 |
15
More hair + face warp continues. Also love the map!
|
I was not enjoying the tie though, so I replaced it, and then I tried to generate myself a new office shirt because this cyber chest was not very affirming for me. It didn't really work with the same prompt and it kept giving me versions of the blue shirt. So I tried to change the chest silhouette a bit with the warp (non-AI) tool and ended up with this (16):
 |
| 16 |
I decided to pivot the aesthetic from liminal boss to tech worker and prompted the background generator for "Tech Office":
 |
17 (a,b)
|
I chose the second one (17b) because I felt it was more modern-looking.
 |
| 18 |
At this point (18) I am tired of this loud pattern and the skin collar so I went to AI-replace the shirt with the prompt "Tech Office shirt".
 |
| 19 |
Tie-removal worked, and it gave me this 90s blouse with a tech-y logo (19). But again I'm seeing some volume where I don't want it. I feel like it's faint enough for me to question myself at this point, but I know at least that I personally don't like this silhouette for myself. First I went to patch the neck(20).
 |
| 20 |
Then I thought I would try the prompt "Tech Office man shirt" and what do you know.. >:-(
 |
| 21 |
I really questioned myself for a second before trying this prompt, but clearly I shouldn't have. The man shirt results (21) are exactly what I wanted, no question. As I'm writing this I realize that I never tried "tech office woman shirt" because I didn't need to for what I was doing, and kept going from there.
{ note from 2024:
I see now that I just forgot about the gender binary while working on this and this is the obvious result..sorrryyyy....🙄 }
expressing emotions:
 |
| 22 |
Felt pretty frustrated with the pressure to be a tech man when I'm more like cosmos. So I took out my frustration on the office and replaced the ceiling with "Hell"(23).
 |
| 23 |
Not my favourite results (23), so I tried "Office Hell" and was very happy with the Calendar app inspired dystopian ceiling(24).
 |
| 24 |
My hair also kept increasing in layers, styles, and volumes! I added a "Boss Crown" to celebrate(25).
 |
| 25 |
smoothie break & back to work:
I thought I would get back to reality a little and changed the
ceiling to "city" and got this sleek subway car feel which was nice.
Tech people love smoothies so the PicsArt AI served me up this beverage,
with writing that went outside the margins which I erased until it
conveniently said "ice"(26).
 |
| 26 |
 |
27 The smoothie grossed me out so I took things in a different direction by
removing it and layering an AI 2.0 generated "animated effect"
transformation filter and combined the two render styles to get this
semi-animated sims-core look.
|
These were some of the animated me's generated. (28, 29)
I love the outfits!
 |
| 28 |
 |
| 29 |
After adding and layering another image of my face onto my post-AI face, it looked even more cyber-femboy (30), which is what I finished this particular project with.
I'm pretty sure I will use this for my About page, since it mostly captures my brand.
 |
| 30 |
my hobbies:
Ok, I also tried "research library" with "server" and "digital" applied to the books which produced this techno-bookshelf(31).
 |
| 31 |
I also tried "digital lab emo 2009" to see if I could express my love for emo style in a STEM context, but instead got some album covers for the band "Digital Lab"(32).
 |
| 32 |
Incredible, but not what I wanted, so I made my own Digital Lab Emo 2009 experience with the last energy I had for this visual adventure (33, 34).
 |
| 33 |
 |
| 34 |
concluding and reflecting:
Pretty cool features and do reflect the binary pressure and confusion of society that comes up IRL. In the end I remixed the gendered options to reflect something that kind of feels like an AI decided how I should look non-binary :( and sometimes that's how I feel in real life too.
I guess those can be the limits (or enhancements?) of using AI Image generation when making art. When I avoid those features, and make art from images in my camera roll based on my narrative and structure-level tools - I don't encounter as many external biases projected onto my likeness, or any visuals I am creating until I have to think about the platform where I might post them.
Ultimately I feel closest to the most uncanny-valley images this project has produced, because the unsettling qualities reflect the cognitive dissonance of the IRL and URL.
xoxo Lockette
P.S.
questions and thoughts I still have:
🐭 don't seem to have a choice whether or not I'm going to get gender-edited even if I am just trying to include a penguin or a mug in my photo, there seems to be a gender-assignment feature included in the AI-generate
🐹 questions on iphone editing and other ways you could be gender-edited in contemporary software with ai features - what are the implications? - how is this going for adobe? does this show up in other ai image generators?
Update from November 2024
Since writing this I have come to learn more about AI and recently did a little research on this gendering situation.
Here are some references I found on this subject:
How Computers See Gender: An Evaluation of Gender Classification in Commercial Facial Analysis and Image Labeling Services MORGAN KLAUS SCHEUERMAN, JACOB M. PAUL, and JED R. BRUBAKER, University of Colorado Boulder
https://www.youtube.com/watch?v=pTSEViCwAig&ab_channel=DEFCONConference
(AI and Security)
#ai #security
Hacking the Cis-tem
#trans #ai #bias #uk
https://ssir.org/articles/entry/when_good_algorithms_go_sexist_why_and_how_to_advance_ai_gender_equity
