I think
Photo Seeds has a fundamental flaw:
Odds are much higher that someone will remember a photo or a number of photos, and the order of those photos before they remember 12 random words.
If you've ever played the memory game (kids are much better at it than adults), you'll know you can barely remember a few pictures. And every new picture you see makes it more confusing.
Well, one can easily have a few super-memorable photos he wants to turn into wallet. He won't have to remember how the photo looks like.
Like "the main photo from my wedding", "the first photo with my boy" and so on. But..
your point is valid because nowadays one may have 20 photos with the first day of his kid, and then the things indeed get confusing.
---
Also, in case nobody noticed:
How can we make idea even better?
Use bitmaps instead of files
Imho the current implementation can be considered flawed, since one may want to print the photos that make the seed, for example, for remembering easier.
(Of course, it can be seen as a feature too, since one may want to do this from certain historic files - the images of his first few floppy disks, or the first version of whatever application and so on)
Back to the images: if the implementation gets corrected/improved as the dev wants... I am not so sure if a 4k photo and its scaled down to 2MP version would return the same seed.
---
Conclusion: it
can become an interesting project, but it may be better to not be linked to the idea of photos, it may be better with (any) actual files. It avoids confusion and other possible problems.