UnlockFile
📚 Technical analysis · 2026-05-03 · 10 min read

BIP39 partial seed recovery — what's actually possible

Honest about what's recoverable. 1-2 words is trivial. 3-4 viable. 5+ is mathematically impossible. Here's the concrete table and why.

Your BIP39 seed (12 or 24 words) is the only thing standing between you and your wallet funds. If you lost it partially — you remember 10, missing 2; or you misread the paper and doubt 3 — the key question is: is it mathematically recoverable?

This guide answers with numbers. No marketing.

How BIP39 works (short version)

BIP39 is the Bitcoin Improvement Proposal #39 standard. It converts a random sequence of bits (entropy) into human-readable words from a fixed 2048-word dictionary (English, Spanish, Japanese, Italian, etc).

The checksum is the magic: tells you if a proposed seed is mathematically valid (15/16 random proposals fail checksum). That reduces search space 16x when only the last word is missing.

Concrete table: how many words can you lose

Missing wordsSearch spaceTime on our clusterViable?
1 word (any position)2048 candidates × 24 paths< 1 minute✓ Trivial
2 words (any position)4M × 24 paths = 96M5-30 minutes✓ Trivial
3 words8 billion × 24 paths1-3 days✓ Viable
4 words17 trillion × 24 paths5-21 days with LSTM hints⚠️ Edge
5 words35 trillion × 24 paths~years✗ Impossible
6+ words72 quadrillion+impossible in lifetime✗ Forget it

The "× 24 paths" refers to standard BIP44/BIP49/BIP84 derivation paths (legacy, segwit, taproot, etc) — Bitcoin Core, Electrum, Trezor, Ledger use different paths. Our engine tries them all to find which one your wallet used.

The hint boost

Three types of information that reduce search:

1. You know exact POSITION of missing words

If you remember "I'm missing words 5 and 17", you reduce position space. Without knowing, we'd have to try 12!/2! position pairs × 2048² candidates. Knowing positions eliminates that factor.

2. You remember INITIAL letter or approximate LENGTH

BIP39 words are 3-8 letters. If you remember "I think it started with S", you reduce from 2048 to ~100 candidates per word (the 100 starting with S in wordlist).

3. You know wordlist language

Official BIP39 supports 10 languages. If your wallet was created in MetaMask with EN UI, almost certainly EN wordlist. Mycelium in Spanish = ES. Confusing wordlists is cause #1 of "my seed doesn't work" (words exist but SHA-256 checksum fails).

When we CAN'T do anything

5+ missing words: 2048^5 = 36 trillion × 24 paths = 864 trillion attempts. Our cluster at 100M ops/sec needs 274 years. If someone says "guaranteed recovery of 6 missing words", it's a scam.

Without knowing wordlist language: multiply everything by 10. What was viable becomes marginal.

Without idea of derivation path + non-standard wallet (Cardano, Solana, etc): we may finish the AI Scan successfully but without knowing which addresses to check for validation.

Real case (anonymized)

Client inherited a Trezor. The seed paper was partially burned — legible 10/12 words + partial of 2 missing ("c..." and "...mar"). Said wallet should have ~2 BTC from 2017.

What to do if your case is "viable"

  1. Make an honest inventory: how many words missing, which positions, what hints (initial, language, approximate length)
  2. DO NOT import partial seed in any online wallet — some save locally what you input and are hack vectors
  3. Find original device (Trezor/Ledger): has metadata of derivation path
  4. Check blockchain explorer with any address you remember — confirms wallet has funds
  5. Pay USD 35 Diagnostic: we tell you in 24h if your case is "viable" or "impossible". Honest.

To start, the Wallet Recovery page has all details. And if you have 5+ missing words, don't spend money on any service — mathematically irrecoverable, no matter who tells you otherwise.

Got 11/12 seed words?

We tell you in 24h if your case is viable. If yes, recovery in 1-7 days with LSTM engine + 5×RTX 3090 GPU cluster.

🚀 Start AI Scan · USD 2,000 →