Dice entropy to mnemonic

I'm experimenting with the idea of generating a new mnemonic for a wallet with dice. I've never owned a hardware wallet (yet) and and wanting to generate a new seed for one in the future.

SIDE NOTE: I'M not going to use the entropy generated that I share here. YOU SHOULD NOT use the entropy example here either. DON'T use the websites with an online computer. KNOW what you're doing before you use any information from this thread.

Back to the question. I bought dice from the dollar store (yes, just testing things out right now) and rolled all 15 in the package a few times and generate entropy. After 99 rolls (this is what I read in a few places, I looked for websites that converted the entropy to a 24 word mnemonic.

3653462546461144335434522636454342256326513252111434324356415313433346212641323141661614122625351213 (DO NOT USE THIS YOURSELF!!)

I pasted this into various sites:

• https://getcoinplate.com/bip39-seed-phrase-mnemonics-generator-offline-online-tool/
• https://bitcoiner.guide/seed/#/seed
• https://bitcoinistrue.com/tools/dice-to-seed/
• https://github.com/veebch/Bip39-Dice (more manual, printed PDF)
• https://github.com/RushabhParmar12/dice-to-seed/blob/main/dice-to-seed.html

The first word of the mnemonic that is generated by the above 4 are as follows, in order:

• 640 - exotic
• 439 - cycle
• 61 - always
• 632 - exclude
• 632 - exclude

The first one, I used advanced mode, turned on entropy, chose dice 1-6 and pasted it in. I noticed that it changed the entries under "Filtered Entropy" so all 6's are 0's. Then a 3 became 1, 0 became 1, then 5 to 0 and the next 3 is 0. I tried changing the mnemonic length and PBKDF2 rounds but still didn't match any other generator.

The second said that my "176" bit of entropy was lacking for a 24 word seed, and I needed 256 die rolls... so I continued:

3653462546461144335434522636454342256326513252111434324356415313433346212641323141661614122625351213424633161623322125164332116246665112353222644653362342435442642536413555435524466156533541154654213346523215324224621243621665516654134426356356525456124642 (AGAIN, DO NOT USE THIS YOURSELF!!)

After pasting these into the 4 sites (and manual PDF), I then was given these for 1st words:

• doctor
• issue
• uniform
• exclude
• exclude

So, the first 3 are calculating very differently.

However, when I take the HEX from the entropy section of the last one (4ef315daeb1456608ee4b5784f166993f11fc6e7e608f908d8d47f6d8d3c955f) and paste that HEX into the first, it gives me "exclude" as the first word. The list then updates to:

• exclude
• discover
• --- no HEX option
• --- no HEX option
• exclude (source of HEX)

I forget where I read this, but the idea of using multiple mnemonic generators with the same dice rolls was stressed to be sure there wasn't something underhanded being done in the code that selected words. If that's the case, why are most of these different?

Either way, with some manipulation, 3 agree on the first word, so those are likely safe to trust. Still I'm liking the PDF manual version.

Does anyone else have experience with this approach and have further input?