Ah, yes, I did see this post. I‘m sorry it was lost in the shuffle after the new year.

Looking back at the comments, I see David Shaw has already covered a number of the key questions. He’s right, physical attributes of the images, like carte de visite or daguerreotype, can be very helpful bracketing dates. Once you have dates, you can narrow down the names to put to faces.

Experts like Maureen Taylor have been working in this area for years. Maureen runs a blog and podcast as The Photo Detective (maureentaylor.com) and has some terrific books out there readily available at the usual sources. Similarly, Lisa Lisson (lisalisson.com/category…) has a number of online talks and blogs on these topics.

In our family history communities here on Substack, my go-to person would beKathy Stone. She’s been a photo organizer and archivist for decades and may have some insights for you about the physical prints. (Tip: Join us for Projectkin’s Kathy’s Corner next week 😉. See projectkin.org/events.)

Regarding AI and facial recognition, we rolled this core technology into our now-offline family photo storytelling platform, Ponga. From my experience in development, testing, and support, I assure you AI is good at this. It uses facial “geometries” to sort through large pools of photos, grouping those you might not have realized were all the same person.

Unfortunately, it’s not great at a handful of photos — at least not with any confidence. Your best bet is to …

1. Start with what you’re sure of (birth/death dates of the possible candidates in the collection of photos, for example.)

2. Then sort through the physical attributes of each image (as David suggested, and you’ll find details of them in the books and posts from Maureen and Lisa) to try to find the date ranges for those types of photos (See the bullets at the bottom of Maureen’s post here, for example, maureentaylor.com/reliv….)

3. Look for “anchors” you can be sure about among the photos or even in the details of the photo, like a piece of tech, book, or other dated item that could not have been present before a given date. Cars are great for this, but fashion can be helpful this way, too. (Her dress looks like 1860s to me, but I’m not that expert.)

   And finally…

4. Consider the known age differences between your contestants and the guessable ages of the people in the photo. Children are golden for this. A two-year-old then will look pretty much like they do today. It’s hard to confuse them for grandparents. 😉

If all that doesn’t get you going, I would try a trial account with My Heritage. From what I know of the AI, this PhotoDater feature has a chance to be worthwhile.

I understand that it pulls from the same variables to find a timeline but does it at AI scale. Fashion and tech are the most compelling elements. Take a look: (blog.myheritage.com/202…)

I promise, I’m not an affiliate. In fact, I’ll warn you. They can be pretty aggressive in following up on trials.

😮‍💨 Sorry this is long. I invested my heart and treasure in making it work as a business. I hope it’s helpful to you now.