How It Started...
You may have noticed the #peanutslibrarians hashtag on Twitter recently. I am not sure who first used the Microsoft Designer Tool to create a Charles Schultz style Library-Avatar, but the first person I saw sharing theirs was Lauren Mobley (who hosts a killer podcast BTW - go check it out!). The idea is simple, just ask the AI to generate a likeness of Librarian You in the style of a Peanuts character. Then, if you're so inclined, share yours via the hashtag.
I'll be honest, at first I wasn't super excited about this trend; I'd seen all the Funko Pop avatars awhile back AND I'm generally very skeptical of trends that require we share personal information with tech companies in exchange for avatars and badges. (I mean, c'mon y'all. All these folks that are _____ certified or who have become an ambassador for ______ are also unpaid advertisers for those billion dollar companies, but I digress). However, I've been chewing on a library+Lego project for awhile now, (more on that later, maybe), so I decided to jump on the bandwagon... but make it Lego!
Of course, it didn't stop there. Here are a few of my creations inspired by different pop culture references.
I experimented with several different prompts with various levels of success. Here are a few things I learned or had reinforced by this process, some of which feel important.
All the images I generated defaulted to white. I tested various prompts, dozens and dozens of times. While none of those prompts included instructions related to race or ethnicity, not a single person of color was generated by the AI.
All the images I generated defaulted to thin body types. It was only when I added the phrase "plus size," when I asked it to generate a character in the style of RuPaul's Drag Race, (because of course I did!) that I got someone with a curvier body type.
I attempted to make images in the style of both "Beyoncé's Cowboy Carter album cover" and "a Taylor Swift album cover" but both requests were kicked back because they violated Microsoft's responsible AI use guidelines. Honestly, I was delighted to when this happened, but also wondered why copying the styles of Peanuts, Lego and Rupaul were cool, but Beyoncé and Taylor were off limits.
These AI generators still struggle with text. The most challenging part of all of this was getting the darned thing to spell "library" correctly!
The Microsoft Designer Tool/AI Image Creator is free, but does require that you create an account. Note: MS did not sponsor this post. Nor did they ask me to write about them/this tool.
How It's Going...
For awhile now I've been doing this activity with groups (including my wonderful Rutgers students - shoutout to RU515!) as a low stakes way to introduce them to what AI can do with images.
During a workshop, I show participants how to use the "I can't draw" feature in Padlet to create an image that introduces themselves to the group. This activity was inspired by something similar shared by Carl Hooker on Twitter and has proven to be an amazing conversation starter. Once they've finished creating their intro images, I prompt them to generate designs related to sports and the arts. Then I ask what they notice about the products they've created. Here are some things we've discussed as a result of this exercise:
When asked to generate an image featuring a basketball player, the AI consistently only generated images featuring Black men.
Conversely, when asked to generate images of ballet dancers, all the images contained white women.
Once during a workshop I was leading in Missouri, a librarian began asking the AI to generate images of people "in the ghetto" or "in the hood." All the people in the resulting designs were people of color. Then, when she asked it to generate images of rich people or beautiful people, all the designs featured white people.
These activities, and subsequent conversations, can and should be replicated with students.
Padlet (which has been around for eons and continues to be wonderful) has a free account version, but it limits users to creating only 5 boards. Note: Padlet did not sponsor this post. Nor did they ask me to write about them/this tool.
What's Next?
Lately, I've been thinking a lot about the Bookelicious Bookmoji as a tool for using AI to help connect kids to story. Unlike the other two avatar generators I mentioned above, the Bookmoji operates, at first anyway, like an old school avatar generator. Rather than writing a prompt and allowing the AI to generate the avatar, users select from a zillion (approximately) options to create an avatar that represents their identity and/or things they love. Here's mine:
This is when the magic starts.
As readers create their Bookmojis, the Bookelicious AI generates a custom bookshelf full of titles based on the choices a user makes when creating their avatar. In other words, as kids craft an avatar reflecting their love of space, cats and the color pink (as an example) books related to all of those things begin to appear on their book shelf. Related subject headings appear under the bookshelf, so kids can expand their search. AND users are allowed to make multiple avatars to reflect how they might be feeling on any given way. Did I mention that it's magical???
Not only is this super fun, but I keep thinking about how powerful this tool would be for kids who KNOW they love sports, Lego, baking, painting, video games or something else, but don't YET know that they also love books! While this is certainly a branch from the same tree as the other AI Avatar generators mentioned in this post, to me this application has real potential to affect our ability to connect readers with books that might just be the one that transforms them into a reader. And, I won't lie, I'm super excited about that.
Creating a Bookmoji is free, but requires you create an account. As with other tools, there's student facing options for Bookmojis that protect the data collected about them and that allows teachers and librarians to see the book wishlists kids may create as a result of crafting their Bookmojis. As an FYI: Bookelicious also offers library processing and accepts POs. Note: Bookelicious did not sponsor this post. Nor did they ask me to write about them/this tool.
A Look Back...
Finally, I couldn't create a post about avatar generators without giving a shoutout to the OG Avatar generator blog post from Gwyneth Jones. Not only has Gwyneth's work always been an inspiration to me, but the title of this post is a direct nod to hers! It's hard to believe that her original is 12 years old now and that we've all been thinking about how to use avatars in our work for over a decade. I guess it's true what they say: everything really is a remix. That said, as I begin to lean into my status as a Library Elder™️, I'm excited to see where AI takes us next.