Wider Lens: A Digital Transitions Podcast

EPISODE 3: Michelle Gollehon, Utah Historical Society

• Digital Transitions

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 48:28

EPISODE 3: Michelle Gollehon, Utah Historical Society

AI Meets the Archive: How Utah Historical Society Is Using Google Gemini to Unlock 118,000 Card Catalog Records

What happens when a digitization project is too big for any human team to tackle and AI becomes the only practical path forward?

In this episode of Wider Lens, we sit down with Michelle Gollehon, Digital Asset Specialist at the Utah Historical Society, for a fascinating conversation about one of the most ambitious and innovative AI-assisted digitization projects in the field right now.

With the Utah Historical Society preparing to move its collections into the new Museum of Utah, Michelle and her team faced a pressing challenge: an entire legacy card catalog — 118,000 physical index cards — that needed to be digitized and made searchable before the move. Captured on DT's Versa imaging system and transcribed using Google Gemini Pro, the project offers a real-world blueprint for how AI can unlock research resources that would otherwise remain inaccessible indefinitely. By Gemini's own estimate, the approach saved approximately 4.5 years of manual labor and $176,000 in staffing costs.

In this episode, we unpack:

  • How a collection move to the Museum of Utah sparked an unexpected AI pilot project
  • Why AI transcription is fundamentally different from traditional OCR
  • The prompt engineering journey: from early hallucinations to asking Gemini to write its own prompt
  • The practical limits of working in batches and other hidden bottlenecks in AI-assisted digitization workflows
  • The tribal flags project: digitizing the flags of Utah's eight Indigenous tribes for museum display using a Mylar sandwich technique on a wide-format scanner
  • And much more…

Whether you're managing a card catalog of your own or just beginning to explore AI's role in your digitization program, this episode offers concrete lessons from someone who figured it out by doing it.

🎧 Listen now for one of the most practical, on-the-ground accounts of AI-assisted digitization you'll find anywhere in the cultural heritage field.

And as always, we welcome feedback, comments, and questions. Let us know what you'd like to hear on this podcast, or if you'd like to be a guest. Email us at: widerlens@digitaltransitions.com

Additional Notes from this episode:

SPEAKER_00

Welcome to Wider Lives, a podcast from digital transition. This series, we sit down with leaders from across the cultural heritage community to talk about the real work of digitization, what's happening today, what's coming next, and what it means to the people behind the project. Each conversation explores heritage imaging in practice, the technical challenges, evolving standards, and complex collections. But we also focus our lens on the human side, the stories, lessons learned, and moments that don't always make it into the conference papers or imaging specifications. Whether you're working in a museum, library, archive, corporation, university, or research facility, Wider Lens is a space to share experience, spark discussion, and think critically about the future of cultural heritage digitization. Let's get started. So hi everyone. Thank you for joining. Another episode of the Wider Lens. This is the first podcast, as far as I know, dedicated to cultural heritage digitization. A little bit about digital transitions. First of all, I'm Patrick McDonough. I'm the head of marketing for DT. And for those that aren't familiar with DT, we're a leading provider of digitization solutions for the world's top galleries, libraries, archives, museums, corporations, and institutions. From turnkey digitization solutions, advanced imaging systems, and sophisticated software, to DT's State-of-the-art Service Bureau Division and our unparalleled training, service, and support team. For over two decades, DT has been trusted by the most discerning organizations looking to achieve the highest level of imaging fidelity and efficiency because we believe every detail is worth preserving. And that is the end of the sales portion of this podcast, I promise. I'd like to get started by introducing our guest. Today's guest is Michelle Golan. She's the digital asset specialist at the Utah Historical Society. So just to kick things off, Michelle, good to have you here. We appreciate you taking the time to join us. Do you want to introduce yourself and share a bit about your background and your current role?

SPEAKER_01

Thank you. Yes, so I am the digital asset specialist at uh Utah Historical Society. Um I've been in the position coming on about five years now. Um I am in charge of all in-house digitization and vendor-related digitization for the state of Utah, um, the historical collection. So it's all the material that um has once been owned by somebody else. Anything government related, that's a different department. That's the archive. So this is only um personal journals and photographs and those kinds of collections, um, the personal collections. Um I got into this profession, I guess that I'm kind of jumping ahead, but I uh was in academic libraries and I was doing uh article scanning for interlibrary loan. And that's kind of how I got started doing digitization. Um, I was using uh a kick, I think first generation kick to do it. Um and my library wanted to do some more output, so we got some equipment and I kind of led the program. Um and I was there for about 10 years, went to grad school for museum studies at Syracuse University, um, back to the original library for another couple years, and then um came, made the move down to Utah and was working at the University of Utah, Marriott Library, for a couple of years before I um started here at the Utah Historical Society.

SPEAKER_00

Amazing. I I'm curious, you know, I don't think many people grow up thinking I'm gonna digitize historical documents for a living. But were you, you know, were you always sort of drawn to museum, historical, those kind of those type of topics?

SPEAKER_01

Yeah, my like probably half of us here, my undergrad was um art history.

SPEAKER_02

Yeah.

SPEAKER_01

So the history, the visual. Um I dabble also in all kinds of arts. So I think it really kind of um was spurred by you know a study abroad programs and um just kind of snowballed from there.

SPEAKER_00

So yeah, that's interesting. Well, what was your medium of choice, or what do you what do you prefer when you're doing art yourself?

SPEAKER_01

Um recently I've started quilting. So you can probably see in the background.

SPEAKER_00

Yeah, I see that nice.

SPEAKER_01

So I've been using that, but I've I've seriously done everything um from metalsmithing to watercolor, oil painting.

SPEAKER_00

Um, I appreciate yeah, sharing some of that background. Um, I think a lot of people listening to this podcast or joining us here on this event will have similar stories about coming through, you know, our history or or some other way and then finding themselves in a digitization department. Um, I want to jump ahead a little bit because I know you've been working on this this kind of large project using AI to deal with metadata and really, really fascinating. Um, I've heard you talk about this before, but I'd love for you to just give us an overview of the project itself and then maybe update us on where you're at with that.

SPEAKER_01

Sure. Um, so it's um kind of spurred, a project spurred by our um move to the Museum of Utah that is going to be opening in two months. Um we are having to downsize our engagement room, which is our um public access research center. Um we've had these really nice old school um card catalogs in our spaces before, but now we have to downsize and have no room for them. Um and it's just more of a research asset. It's kind of one of those things somebody spent a lot of time doing this, and it kind of indexes um our resources by topic and by um um biography. So it um helps the the who, what, where, and when if we can't find it immediately through catalog searches. Um we're kind of in this holding pattern because we're packing the collection, moving. Um, and so it felt like a good time to do this collection, especially because we got our um DT Versa in July of last year. So um that's what we did. We started this um hard catalog digitization project, um, all shot on the VERSA, which was really speedy. We got that done. That's been done for months now. Um and now the hard part is you know making it a usable system, somehow how to make it accessible and searchable. Um, so we decided to do some testing with um Google Gemini. Um the state of Utah has um uh enterprise contract with Google Google. So we have Google products, and so I just started messing around, throwing some PDFs in there, some TIFFs in there to or and JPEGs um to see what worked best, kind of messing around with it to see if it was feasible. And um sure enough, it was, and um we have done that for all 118,000 cards. That's what it came out to be. Um, so quite a few cards. Um, they've all been digitized right now, they've all been thrown into Google Gemini to transcribe. Um, the majority of the cards were um typed, so pretty easy for um transcription. There were a few with notes and um pretty elaborate cursive, and they did Gemini did a really good job with that. So um right now I am uploading into the DAM our digital asset management system, which is hosted um by the University of Utah Marriott Library. Um, and so just kind of chipping away with that while we wait for the collection move and the museum opening. And um so yeah.

SPEAKER_00

I wanna I want to dive in a little bit more because um like OCR has been around for a while, but what you're doing is is more than just typical OCR. So I'd love I'd love to know one when you discovered that AI would actually be useful for this instead of just like, you know, here's an interesting tool, will will it help us? Uh, what that was like and then what you're actually doing with this that goes beyond just standard OCR.

SPEAKER_01

It's doing smart stuff behind the scenes, which that's you know, kind of nice. That so I did some testing and you know, just making a PDF of an image and having the PDF do the OCR through Adobe PDF OCR versus the transcription, and it was night and day different. Um just the subtlety that Gemini was able to um pick up on, and um, especially because this is like a series, so um, it would pick up and learn things, but you know, uh just a plain old OCR wouldn't.

SPEAKER_00

Because the OCR is just gonna try to read each individual character where the AI is gonna have a little bit more context and try to interpret, like, oh, this word is probably this because of the context it's in. So I'm sure it can kind of get over the that block of like sometimes characters aren't, you know, especially if it's a typewriter, the character didn't type, you know, get the ink on the paper properly, those kind of things. So I would imagine it helps out with that.

SPEAKER_01

Yeah. So, and I had Gemini start policing itself. I said, Hey, you know, if you find a passage that's faded or hard to read, flag it so I can go back and read it myself. Um, so that really kind of sped up the QC process, um, where most of it was, you know, just semantics, you know, like spelling this is spelled wrong. I think they meant this, you know, this card's out of um order because we're going alphabetical. Um, so we had a lot of those things, like it was learning things and it was too smart, like it was giving me things that you know I really wasn't interested in. So my prompt is very long. I kind of um organized a prop prompt just refining it as I found out more issues, like it really had a problem with the um number symbol for some reason. And it would always add characters or do weird things with like volume if it was volume number symbol two, it would like do volume 22 instead. It was just really weird things like that. So I I said, hey, pay attention to the you know, these symbols, um, and it cleaned it up really, really well. Um one problem that I did come across was um after my uh part-time employee finished up doing the scanning, I had her switch over to helping me with the um uploading to Gemini and the transcription. Um so we were getting two different outputs then, and they weren't jiving because I had this whole history of doing this, and she was just starting with the history. So it was um a whole different uh mismatch of ways that it was flagging things and catching things. And um, when she would tweak her prompt, she's originally started with my prompt, but she would tweak it and it would just you know output total different things.

SPEAKER_00

Just kind of using different knowledge that the AI has acquired and kind of understanding. I I want to talk more about the prompt and how you got to that in a minute, but um I want to start with when this project started, you know, you you mentioned 118,000, you know, cards you've got to scan. Was using AI, was it one of those things like this is this is gonna be a big help, or is this where we can't do this without AI? Like, how would you have accomplished this without the AI tools that we have now? And how long do you think that would have taken?

SPEAKER_01

Yeah, I don't think we would have tackled this problem if we didn't have access to the Gemini. Um I think it was kind of one of those things where they were looking for test cases of how would you use this within the the Utah state government? And um I was kind of on a test group for that. And I I used it for some some other metadata, some translations, that kind of thing, some, you know, just really small scale one-off things to test and see if it would be uh useful for us. But this is, you know, pretty straightforward. We aren't asking it to think, you know, and give us, you know, poetry or anything like that. We're just having it transcribe what's on the card.

SPEAKER_00

I think a lot of people is, you know, especially in like the artist and artist adjacent sort of worlds, the idea of AI has become like a dirty word because you think of it like, you know, it's creating something or you know, taking away a creation role from an artist or something like this. But in this context of like it's it's unlocking an ability we would not have had just from data processing and interpretation. I mean, without paying someone to sit there and just manually transcribe 118,000 cards, like it's it's an almost impossible problem to really tackle and make use of.

SPEAKER_01

Going on that, I actually had Gemini kind of calculate how much time we saved um doing this project. So after we had everything digitized and we had some numbers and um some times, you know, like how many hours it took to do this, how many hours it took to do that. Um, I had Gemini kind of calculate it, what it would take for the whole project and how much we saved. And it says that we probably saved about four and a half years um by using Gemini to transcribe versus hiring someone to do it. So yeah, and it was like $176,000 that we saved. So wow, that's incredible. Love the time saver. So that's yeah, I don't think this project would have happened unless if not for this.

SPEAKER_00

That's such a it's such a great story, too. I mean, uh, you know, AI obviously is just such a hot topic. Everyone's talking about it, but I think a lot of us are still trying to figure out how to use it because it has all this potential, but how do you actually apply it and make it a useful tool? I think this is just such a great case study. So backing up a little bit, I want to talk about your discovery and kind of exploration of this process and where you've arrived. You mentioned something to me once about the prompt, and you got to a point where you actually asked the AI, like, what do you want? Like, recap that for me. I'd love to hear that again.

SPEAKER_01

My, you know, first prompt was like, actually, I kind of kicked myself for it because I asked it to summarize the image or the image, which was the image of the card. So it was text. Um, and for the first couple of times it was doing what I wanted to do, was it doing a transcription of it? But then when I finally um added like a whole PDF of a hundred um images within the PDF, it was kind of doing a summary. And I'm like, okay, that's you know, I made a mistake by asking a summary. I want a full transcription. Um, so I just kind of you know saw what it gave me. I asked it what it can do too. So um, these are the things that I can do, and I'm like, oh, that's what I want you to do. So I told it, and then after I don't know, I had a couple good outputs and I got a uh a system going, I am like, how would you want this prompt? How would you write this prompt for what I want? And so I outputted a prompt, and I kind of that's the basis for my my prompt now. Um, like I said, I would tweak it and add things to it. Um, but after you know, like I said, it learns what you want and what you're doing, so it kind of anticipates a little bit, I guess. I hate to say that.

SPEAKER_00

I I think that's just it's such a fascinating thing. And I've I've done that myself. We're trying to, you know, figure out like, oh my god, how do I ask for this and get what I want? And literally just said AI tool like Chat GBT, like here's the goal I want to do. How would how would you like the prompt structure for that? And it'll spit it back out. And you know, you might have to adjust some details here and there, but um, it's a it's a really useful tool I think a lot of people might not think of. So I just love love hearing that story. And you know, you mentioned like you start with this prompt that it's sort of seeded, and then you've I'm sure evolved that. And what what does that evolution of the prompt look like? You mentioned that you have like a really solid prompt now. What has that process been?

SPEAKER_01

Um, so I've been using sticky notes on my computer, and I have, I think, three or four of them dedicated to um this project and prompts. Just the the tedious work that takes all the time um doing the back and forth thing that I've that's kind of what I've used it for in other cases.

SPEAKER_00

And how how far into this did you get before you you fully trust the output? Or do are you still sort of keeping an eye on what's coming out of it?

SPEAKER_01

So the hallucinations were really prevalent at first or getting errors. I kind of errors and a lot of hallucinations. And um I originally started using um the Gemini for the workplace with his built into Google Sheets or Google Drive, um, and doing that. And I really quickly learned that that doesn't have a lot of computing power and it's very limited in what it can do. So um I went to the browser version of Gemini, which is Gemini Pro, and um used it there. And it was a, you know, it opened up a lot more capabilities, less hallucinations. I asked it like how what size of PDF is a good PDF size for you? And so it said about 100 pages. So we limited our output through the Versa to 100 pages. Um and it has a file size, the file had to be under a certain size. So um it all had these limits. So that's kind of where you know, I just worked within that. I asked it what the limits are, then I went within it. Because in in trial and error, I was just like driving myself crazy. Why isn't this working? But this does sometimes. So it was going back and forth. Um, sometimes it would work, sometimes it wouldn't. And so I finally just got this, you know, down a hundred images in a PDF, upload that.

SPEAKER_00

Yeah, I mean, it's a it's a whole whole whole different um QA process for AI. Like you talk about catching hallucinations. I mean, that's different than your typical, you know, color accuracy sort of check. Um, what how is that how has that been? How did you handle that?

SPEAKER_01

Well, it eventually, you know, I was keeping things in all one chat because like I said, it learns from each other, it's learns from itself. So I was trying to keep things in a chat and the chat had a limit that I found out. And the further down the limit, down further within the day of the chat, um, the further and the more crazy hallucinations I got. So I learned to do, you know, shorter chats. So, you know, maybe like all the A's in one chat, you know, all the B's in another chat instead of trying to do all of them in one chat. Um, because I don't know. Like I said, it just kind of like the more you ask it to do, the more it gets confused. And when you say, hey, look at this PDF, it's looking at the PDF like maybe five conversations ago or something like that.

SPEAKER_00

It's funny, I've I've had the same issues with with um Chat GBT as the one I was using at the time, but um, and just having to like, okay, we're just gonna reset and start a new thread here and not try to just kind of overfeed it in there. Um something I'd like to hear about, like with all this data, and you mentioned Google Drive, you know, where where are the bottlenecks here? I mean, you're moving big files around, you're you're capturing images, you're outputting the data. How are you managing managing that? And what are you doing with that data as it gets moved around?

SPEAKER_01

Um, that's been a real stress point for me and my team. Um the file large files and um moving them back and forth from local to Google Drive and then to my server for the dam. Um I actually realized that the building that we were in, it was a repurposed building that houses temporarily because we were in the middle of this move and everything like that, it did not have high speed internet. It was kind of um really, really um ancient speeds. So um that was the big hang up. And finally, you know, I would just start doing things from home where I have fiber. So um that kind of speeds things up. Um I put a ticket in for um them to replace the, you know, check just check the network. How's the speeds? And they're like, yep, this is outdated. They admitted it, and um, it was kind of funny. The day before we moved out, it got fixed, honestly.

SPEAKER_00

It was the day before. Perfect.

SPEAKER_01

So that didn't help us. Um, so yeah, that was really a bottleneck, was uploading, downloading, um, just things like that. I don't know if it's just me, but I know we were using Adobe products that's for creation of um PDFs. And for some reason, it doesn't let me like run two different programs at once. So I would get freezes and lose work and yeah.

SPEAKER_00

And I I know, you know, Adobe, when you say about PDFs, in my experience, they tend to be just bulky, kind of fat files to begin with. I actually use a site for anyone that that runs into this problem. If you go to ILovePDF.com, there's a PDF compression free tool that you just upload your files and it compresses. Done. It'll save like 90% of the file size and keep the quality. Um, for those kind of things. I don't know if that would have helped your project at all, but it's those kind of issues can can really add up when you're just waiting for like big 10 megabyte PDFs to to you know be uploaded or downloaded.

SPEAKER_01

Yeah, so that's kind of stuff that I would have like probably investigated more, you know, in hindsight now that it's we're coming on the tail end of the project. I probably would have figured out how to do mass creation of um PDFs.

SPEAKER_00

Yeah, those kind of things. And you know, anytime you you're you're sort of like on the vanguard of a new process and a new way of capturing data and processing. And so I I can't imagine how much you learned through this journey. And like if you started over again, how much different this would this would be. When you look forward to, you know, developing this process and going through this project, do you see ways to apply this type of I guess sort of data capture and process from images to other projects in the future? Like, you know, recognizing faces instead of words or other kind of metadata.

SPEAKER_01

Um yeah, I you know, I really want to try the faces. Um, I know like Google on your phone can group your photos by person. So it already does that kind of thing. Um, even does it for my cat. So why can't it do it for people? So um in the midst of everything else, I was doing a digitization project for um our local group and just a massive output, both slides and negatives and prints, just kind of a lot of I forget, I think it's about 6,000 images. Um, and I know there are duplicates in there, so that's another thing that I would like to see if AI can help with was between the different, you know, I have a print and I have a negative. Um, can they can go through and you know, tell me which ones are duplicates? Um, and then once we get some information about who's in these photographs, see if it can help with identification.

SPEAKER_00

Um have have you shared this with with peers? I would imagine other people in the industry you talk to would be really interested in hearing and kind of, you know, tell me how you did this because we have a similar issue. Like, have you run into that much?

SPEAKER_01

Uh no, I haven't yet, because like I said, we have all this other things going on. Kind of baby steps first. Like I said, I it's digitized. I haven't even started like uploading it or cleaning up or anything like that. So it's very messy right now.

SPEAKER_00

Yeah, I love it. I definitely see, you know, maybe even after this podcast, you might get some people reaching out um for some advice.

SPEAKER_01

Please do.

SPEAKER_00

Yeah, and speaking of which, I just kind of throw it out there. Like for other institutions that might have something similar or maybe just want to start incorporating AI into their digitization program um and you know, related to data capture, those kind of things. Where do you recommend they start or or what advice would you give them as they start looking to build that out?

SPEAKER_01

Yeah, I don't know. I was someone actually emailed me this question um to see if this would be of help. It's really hard to know because each project, all of our projects are different, you know, they're all so unique. Um, so it's hard to give just general advice.

SPEAKER_00

Check your internet connection, make sure you have high-speed internet would be a good one. I think you know, we're so used to just using our web browsers almost just like a window, you know, into something local, but knowing you're uploading anytime you put something in a an AI chat, you're uploading it to the internet and waiting for it to process it and download again. So those those speeds make a difference. Um yeah, I I think you know, just things you might not think about is you know, the the internet speed and how big of an impact that can be when you're scanning 118,000 objects. And you know, yeah, I thought it just really interesting, just kind of recapping the um asking the AI robot, like how do you want this? This is what I want to do. Tell me how you want this. I think those are just really intriguing points there. Um so let me ask you this. Where you're standing today and looking at this project kind of through through the other side of it largely, what's one investment, you know, not just technology, but maybe process people that you think had the biggest impact on this?

SPEAKER_01

I just think it was, you know, the access to the the software, you know, like Google Gemini, I know like that's one of the things that just having access to it, maybe just like research things that you might need in advance. Like when um speaking of the the duplicates, Gemini has an extension or Google has an extension that you can add. But because I'm with the state of Utah, I have to get permissions to use anything new that's not been approved. And I asked for that one and they didn't approve it. So uh maybe having a secret computer that you can download and use as like a testing, like an offline tester for Did they give you a reason why that seems like such an odd denial? No, I I I put in the ticket and I it was kind of crickets, so no one really responded to it. They're probably like, This girl's crazy.

SPEAKER_00

Yeah, I think you're running those things where you know the person on the other end doesn't even know the tool and it's too new and they don't want the responsibility of saying yes and then it, you know, creators or some data leak or something. Well, I really appreciate um that whole project is just really fascinating. I think for many of our audience, you know, kind of experimenting or curious about using AI tools, it's a great case study of ways to to leverage that and kind of hearing the journey there. Uh, I'm gonna switch topics now. I kind of want to just just talk to you, Michelle, about kind of your experience, your thoughts just about you know, cultural heritage, digitization um in general. Is there anything in this realm that you're excited about right now? I mean, obviously AI is a big one, but anything else in the cultural heritage world, whether it's digitization or just general art museums, those kind of things that that's on your radar of something?

SPEAKER_01

Yeah, well, we just got this fancy new tool that will enable us to do a lot different different things. So we have a couple of big projects that we're thinking about doing with the Versa. Um, but I, you know, I got a taste of 3D scanning. We did um one of our larger museum objects, we got 3D scanned um because it's it's massive and it would give people an opportunity to get up and close to see it. Um and it it's it's it turned out really well. Um and just having that I don't know, I don't want to say bright and shiny thing because but because it is it's a car, so it's bright and shiny, it's got crumb. But um it it's you know, I'm such a visual person and seeing the whole thing and being able to, you know, see different angles and kind of see how it just it I don't know. I like that, you know, I like being able to manipulate it and seeing um all sides, seeing the bottom, because a lot of people don't see the bottom when it's a museum or a stand or something like that.

SPEAKER_00

So that's really interesting. You're actually the the second podcast guest who's mentioned 3D scanning as something that they find really interesting. What what was the process? I mean, there's photogrammetry is kind of the what's been around for a while. The new one I've seen is called Gaussian Splats, which is really fascinating. If no one's seen it yet, it's sort of like photogrammetry but kicked up 10 levels. It's really, really fascinating. But what what was the technology you were working with?

SPEAKER_01

Um, so we uh we hired somebody, a contract worker, to do it, and it was the um laser lidar, I think. Oh um, and I I think they combined a couple different methods. Um, and then they added skins to it, they got color samples and stuff like that. Yeah, um, it was really complicated file when I got it back. So, you know, it's moving forward with the museum, it really presents with some opportunities for new learning, new ways to learn for us here at the Historical Society.

SPEAKER_00

Yeah, absolutely. Um, I love it. So, next question for you. If you were not doing this, if you had not found your way into digitization, what do you think you'd be doing instead?

SPEAKER_01

I don't know. I probably would be a starving artist. Like I said, I've dabbled in stuff and I've, you know, I've got people say, oh, you should do this more. You should, you know, really put more time into it. And like I said, I've dabbled if I, you know, really put my mental capacity and my time into it, you know, probably would be an artist.

SPEAKER_00

Uh yeah. So, you know, with the Utah Historical Setting, you're working with a lot of historical material, obviously. Um, do you ever think about how this is going to be used? Like you're basically creating access for things that people have not had access to, you know, a digital record that people from around the world can see or interact with versus the physical object that's so limited. Like, did you ever think about that and that responsibility? I know you recently you had something with um the the flags, if you want to share that, but I'd love to know how you feel about that.

SPEAKER_01

Well, I hope what I'm doing now is still considered good quality in the future because stuff that was done 20 years ago, I'm thinking about redoing some of it because it's just so small. The resolution's tiny, you know. I just want it to be still viable, I guess. Um, so that it is available for the future generations. Yeah. So that project about with the tribal flags, also related to the museum. Um, we have eight tribal tribes within Utah. And so we're giving them space within the museum to kind of represent their the different designated areas. And it um came to our knowledge that they didn't have high-resolution images of their flags. Um, so the director said, Hey, Michelle, can you? And so I scanned their flags. Um, I put them between Mylar and ran them through our um our wide format scanner, and then I turned them into um graphics. Um, so it was fun to just go back and do that and make them really fun graphics and vectors and different formats. You know, I gave them a usable JPEG, I gave them a high resolution illustrator. So it was, and they were really appreciative of because they have something now that they can use in their communities. And um, so it's kind of that old adage, we just don't want to take from them, we want to give back. So it was kind of fulfilling that way.

SPEAKER_00

Yeah, that's amazing. Just you know, something that that's so important and so representative of their culture and their history and their people, and they don't have a digital representation, representation of that and being able to unlock that and create that. It's it's it's wild that they didn't have that. And so just being able to fulfill that is amazing. Um, we're getting kind of to the end here. I'm uh I'm gonna do one of my favorite parts of this podcast. We're gonna do the lightning round where I'm just gonna ask you some random questions and just spit out the first thing that comes to mind. Um I'm just gonna start with this. What's the most challenging format you've had to digitize?

SPEAKER_01

Flags.

SPEAKER_00

That's fair. You mentioned sandwiching them between Mylar. Is that what you said?

SPEAKER_01

Yeah, yeah. Because it was fabric and it was slippery, you know, the flag material. I I don't know what material it is, but it's slippery. And so the the wheels wouldn't grip on it. Um, but if sandwiched it between Mylar, it would feed in. But a couple of the flags had fringe. So little stop and goes. Um, me and my co-worker, we we got it done, and it only took us half a day to do eight of them. So we're really proud with that timeline. Um, yeah.

SPEAKER_00

Yeah, I I love that. I think you know, people outside of the industry think of digitization as like, oh, you just stick the thing in the scanner and you get an image of the other side. But um, this actually leads to my next question was knowing that so much of what you do is is problem solving, with all the high-tech stuff we've discussed, you know, AI or the the Versa and like this 150 megapixel imaging system and all this stuff, what's the most low-tech item in your digitization studio that you use?

SPEAKER_01

I think we have some painter sticks here that we use to prop open pages and stuff like that. We're problem solvers, because we have to scan this really unique item and get a good image. How are we gonna do that?

SPEAKER_00

Sure. I think the the flag is a great example of that. Like, how do you roll this, you know, loose fabric through here? It's a really clever, clever solution. Um, so the next one, and this is this is fun. I'm I'm curious to hear your answer. When someone asks you what you do at a party that you're just meeting, what do you say? How do you describe what you do?

SPEAKER_01

Yeah, I usually just say digitization, and then they try to say that word back to me and they can't quite get it. So um I kind of explained to them, you know, the elevator spiel about, you know, I take, you know, photographs and manuscripts and you know, scrapbooks and scan them and then place them online for access. Yeah, I I think it's it's pretty straightforward what I do, whether or not they get it. It's sometimes a glazed look. I don't I don't even think my family knows what I do.

SPEAKER_00

So yeah, it's funny. You can say digitiz digitization. Yeah, I I I literally as I can't say digitization. Um, but you can tell somebody you do digitization, um, but then you have to explain what it is. So, like, yeah, well, this material comes in and I do it, and this is what's done. What and yeah, people are either they're really interested or they're just like, okay, I've stopped listening. So, last question in this particular round, then we're gonna get to the audience QA. So, for those um in the audience right now, if you haven't any, um if you want to submit a question for Michelle to answer, put it in the QA window, and then our team will help um you know pass it on and we'll let Michelle dive into that. But last one for me. If you could just make wave a magic wand, what's one thing, like a tool, hardware, software, people that would make your job easier or better?

SPEAKER_01

I think it's going back to the tech. The tech is kind of the hang up, whether it's slow internet speeds or a locked computer, you know, just having what I need when I need it. Um, even getting something's tools sometimes is really hard through the state process. So yeah, just having what I want when I need it.

SPEAKER_00

Yeah, faster tools. That's always frustrating when you're like you're just waiting on the computer to do something. You're like, I could be getting so much more done. I mean, I the the days when we can plug a USB cord into our brains and just like process things will save us all some time. Um, well, that's that's kind of it for my question. I'm gonna turn it over to the audience now. We've got some questions that that we've pulled in over the course of this conversation. Um, Michelle, I'll kind of run through these and and love to hear your thoughts. Um, we have Rachel sense Myers, which which hi Rachel, I've I um I know we're very familiar with you. So uh she asks, are you able to batch process cards with Gemini or do you have to do them one by one? You mentioned like PDFs and sort of a limit there. So maybe give some more explanation on what the limits are there. And and it can you just give it one PDF or can you just give it a folder of a bunch of images? How do you address it?

SPEAKER_01

Yeah, I tried giving it a folder and it can't see things for some reason. Yeah, so it had to be a PDF and it had to be under 100 to 100 pages. Um you could get away with like 110 sometimes, but you know, 101 wouldn't do it one time. Um it really is kind of one of those things where I tried Gemini has this thing called gems, and um I once again I didn't know how to use it. It seems like they're more of a thing for um like processes and um things that are you know kind of like a you wanted to apply this to that kind of solutions, and you do it a lot. And I probably would have experimented with that more given the time, uh, because I think that probably could have done some things more in batch, but I tried doing it once or twice and it wasn't giving me any better result than the the normal chat pro was. So um I kind of just let that go. So it was it was really weird, plus because it needed to be formatted in a language, different language. It wasn't semantic. You couldn't just type in the prompt, you had to format it weird. So yeah, it was just something too much for me to do in my time frame. And um I did experiment with that. So yeah, it was very limited, and that's probably one of the the biggest downsides of this project.

SPEAKER_00

I have another another question. You're just talking talking about using Gemini, and you touched on this a little bit in terms of being a government um program, but were there cons um concerns about security with using Gemini and and these things?

SPEAKER_01

You know, um we have a little warning at the bottom of all of our Gemini products that says your state of Utah chats aren't used to improve our models. Um, and of course, the general mistakes happen kind of thing. So um we know that they're kind of an in an isolated bubble. Um, they aren't being used to like they're they're they're in internal, I guess. And like I said, it was just pure transcription. So it wasn't like this material was PPI or um anything, you know, it was bibliographical records kind of stuff. It was it wasn't you know, like, ooh, we're we're exposing, you know, the the state secrets or anything like that.

SPEAKER_00

It was just yeah, I think you know, there's always concerns like you know, maybe that there's this is low stakes material, so to speak, but there's concerns are they gonna take this? Is it gonna pop up, like you said, like in the training data, and suddenly other people are gonna start getting this information in their you know, chat requests. But there are ways to to set up these these systems where they're siloed and that information doesn't leave a specific area and they're more secure. Um change, one quick question here and you know, this is a little self-serving, but I'll just ask it. It's not for me, it's from our audience. Um, but they're asking when you got the Versa, did you do it with the intention of this card catalog project specifically, or did you choose it for other purposes? What was the reason you went with the Versa?

SPEAKER_01

No, this was just something that we could do in the small time frame that we had between moving the collection and opening the museum. We got the Versa because it really opened up the possibilities of a different format. Um, I have before this, I had a large format scanner and a couple Epson flatbeds. Um, so this really gave us more possibility. It has the capability. We got the um the Versa cradle too. So we do we have a lot of library books that are really tight spines. So this will be able to allow us to do um books, bound material, a lot of bound material. We have out of a jillion yearbooks that are on the list to do. We um have started doing our city directory collection, and so that's something that's on the list to um do as well. Um and we have tons of architectural drawings that really have been neglected. So I think you know, we got the vacuum table, and the vacuum table will allow us to do those drawings that have been rolled or folded or something like that.

SPEAKER_00

So um I mean it's it's called it's called Versa for a reason, just versus lots of different material.

SPEAKER_01

Pretty much, you know, it wasn't a hard sell for us. Um and yeah, we we the the future is gonna be negatives. We we got a really big um collection of um film strip negatives.

SPEAKER_00

So we also got the the negative, the film Yeah, the the DT film scanning kit, it's it's a popular one. Um we're guys yeah, we're almost out of time. So I just I want to get to these last few questions because I think we have a few questions around the actually the flag project. I think a lot of people are really interested and fascinated about that. Um somebody asked how you're handling textiles for digitization. I mean, you you you mentioned the the Mylar sandwich, but have you run into other situations where you had to come up with other solutions for textiles?

SPEAKER_01

Yeah, um, when I was at the University of Utah Marriott Library, um we did a couple sashes because it was um 2020 and the all the um this the the the marching protesting and stuff like that, we kind of collected some items and did um a sash too, kind of very similar way. Um and I think there was a pink hat too. Um I've done earrings on a flat lead scanner.

SPEAKER_00

With like the sashes, did you just lay them flat or did you go with the did you have the same?

SPEAKER_01

Yeah, so it was um luckily enough we fed them through the scanner, yeah, through a feeder with mm sandwich in between mylar. Um the mylar really kind of is the key because it it gives it a consistent hole, I guess. In yeah.

SPEAKER_00

Um for the going back to the flags, what were the size of these? How how large were the the larger flags?

SPEAKER_01

Um the largest, so I have the 60 inch wide scanner. So about you know, probably like 55, 50 wide, and then long was probably about 70. But one of the projects I was doing before this card was um digitizing our railroad maps, and um our old location was a historic Rio Grande building, like railroad depot. So we have a lot of railroad maps from that. Um and it was a rolled railroad track pretty much, and it was 80 feet long, so it was really thin. It was like only about um 15, 16 inches wide, but really long, and it was just tightly rolled, so it just was fed in and it's a really large, um, obviously file, but 80 feet long.

SPEAKER_00

So again, you know, we we got a lot of interest on the flags thing. So we'll ask another question on this, and then we might have to cap it at that. I'll see if we can squeeze one more question in. I I want to be respectful of your time as well, Michelle. But in terms of working with the travel groups, have you done other projects with them? Have they asked for other digitization? Is there other historical artifacts that you've worked with on that?

SPEAKER_01

No. So um we have a whole department within the state government. It's one of our sister agencies that are pretty much the only ones that are allowed to work with the Indian tribal relations just because it's a government-to-government entity. So um, but I do and various people have gotten pulled in to do projects with their involved in with that. Um, but because it's a government-to-government thing, we're kind of really um siloed in a way. Um but yeah, I'm open to doing more projects for them because like I said, um we're um the museum's gonna be hosting a lot of their um historical objects and um stuff like that too. So it's gonna be a bigger part of the chip in the future as we move forward.

SPEAKER_00

So awesome. Um very, very interesting. And and I think like I said, there's a lot of interest in in our chat and questions around that. So obviously an interesting topic for all. Um well, I think we're gonna wrap it up. Um, Michelle, I want to thank you again for joining us and sharing your experience, your knowledge, your thoughts. Um, really, really interesting conversation. And I want to thank everyone that joined us today in the chat in the QS QA. Um, be sure to visit our website, check out more upcoming events. We actually have three more live podcast recordings already on the calendar. We have Maggie Wins, uh, she's a lead web developer and digital archivist with the NHL's Washington Capitals, which will be really interesting. That's on May 21st. And that one, she'll actually be here in our office. Uh, we're having a little event here, and so that'll be a live conversation um as well, but you can definitely join us online for that. Uh, we also have Martina Hoffman from the Swiss National Library on July 2nd. Um, she's, you know, obviously Swiss National Library, great institution. She's also um quite the character. I'm really looking forward to that conversation. Uh, we have Sierra Arazzo, she's a curator of collections at NASCAR Hall of Fame. That'll be a really interesting one. There's a lot of AI talk in that one as well. So for anyone that's interested in AI and how that applies to digitization and cultural heritage, uh, we have some really fascinating topics to discuss then. That's on August 27th. We have a couple other events I think would be interesting. We're we're doing basically like a digitization um AMA. Ask me anything if you're familiar with with Reddit parlance. Uh, we're gonna do one in June, June 18th with Ben Quartz. Um, and basically we're just gonna allow people to submit questions both uh beforehand. You could send us questions and we'll feed them to Ben and we'll have a live chat where people can just ask anything you might want to know to some very, very technically experienced and knowledgeable people. And we'll just kind of be an open conversation. Uh so Ben Quartz doing the first one in June, and we have another one in November with Doug Peterson. Um, anyone familiar with DT to any level will will know Ben and Doug's name, I'm sure. All these events are free, they're all virtual. You can find the registration links on our website. Um, and we'll also be sharing an email after this event uh to everyone that registered and joined us with some of these details and also some links to other information we might have touched on during the conversation. Um thanks to the DT staff that joined us. You guys can turn your cameras back on if you don't want to hide any more. But we have um we have Alana Ursia, we have Dan Oster and Daniel Horvath. Thank you guys for doing such a great job with the chat and the questions and managing all that. Really, really appreciate the help. Um, and I'm just the one, you know, kind of voicing things. They're helping out behind the scenes quite a bit to make this all happen. So keep an eye out for the official podcast release of this episode that'll happen probably in the next week on all the usual podcast channels. Um, you can come to our website if if um you need a link to get to that. Um and yeah, until then, thanks for helping us focus a wider lens on the work that preserves our past and shapes our future. So thanks again, Michelle. Really appreciate the conversation.

SPEAKER_01

Yeah, thank you.

SPEAKER_00

Thanks for listening to Wider Lens, a podcast from digital transitions. We hope today's conversation offered practical insight into the evolving world of cultural heritage digitization, both the technical realities and the people behind the projects. If you have thoughts, questions, feedback, or ideas for future guests and topics, we'd love to hear from you. Reach out anytime at widerlens at digital transitions.com. And if this episode was valuable, please share with your cultural heritage friends, co workers, and peers. Until next time, thanks for helping us focus a wider lens on the work that preserves our past and shapes its future.