QR codes are an imperfect solution but probably the best we have for now.
Word list idea is good, but very difficult. You might want to try some kind of character encoding.
The current dictionary uses 2048 words to represent 11 bits per word (2^11) because it's getting close to the limit of how many 3 or 4 letter combinations match recognizable words, making for a good mnemonic device.
There are 17576 possible combinations of 3 letters, but we already have to go into 4 letter combinations to limit it to recognizable words instead of random letter combos.
To put 256 bits in 5 words, you'd need 52 bits per word, so you need 4,503,599,627,370,496 possible combinations per word (2^52).
There are more than that many possible combinations of 12 letters, but not so many words people know.
You're better off losing the mnemonic device and just using different text encoding to make a more condensed format of npub (which is of course just a text encoding format for the binary public key).
That's really hard, so for now, we have QR codes.
Sorry for wasting your time with one of my caveman-esque replies, by the way. I thought I'd take a crack at it since nobody else had addressed the math of your question in a couple hours, even though I understand that the almighty nostr:npub1235tem4hfn34edqh8hxfja9amty73998f0eagnuu4zm423s9e8ksdg0ht5 devs say I'm "not technical," less capable of providing useful input than a chat bot, etc. and they also don't give a fuck if Digit is safe, the woman who inspired me to find math questions interesting enough to try answering you.
Yes that sounds reasonable from what I remember. It might work at around 9-10 words with a plausible dictionary but maybe that is too large to be viable. I'm sure this is a fairly well studied problem.
Thread collapsed
