Encode and decode Base32, Base58 and Base85 without leaving your browser. Base32 here follows RFC 4648 exactly and is checked against the specification's own test vectors. Base58 uses the Bitcoin alphabet — the one with 0, O, I and l removed, because those are the characters people confuse when copying an address by hand — and correctly preserves leading zero bytes, which a surprising number of implementations drop, silently truncating any key that begins with one. Base85 packs four bytes into five characters and handles the z shorthand for a run of zeros. Text is converted through UTF-8 throughout, so accented letters, CJK and emoji all survive the round trip. Invalid characters are named rather than quietly ignored.
| Encoding | Alphabet | Typically used for |
|---|---|---|
| Base32 | A-Z and 2-7 | TOTP secrets, DNS labels, case-insensitive systems |
| Base58 | 58 chars, no 0 O I l | Bitcoin addresses, keys copied by hand |
| Base85 | ! through u | PDF and PostScript, where density matters |
| Base64 | A-Z a-z 0-9 + / | The general default — see our separate tool |
No. Everything runs in your browser. That matters here more than usual, since these encodings routinely carry keys, secrets and authenticator seeds.
It omits 0, O, I and l — the pairs that look alike in most fonts. Base58 was designed for values people copy by hand, such as Bitcoin addresses, where a single misread character makes the whole string useless.
Usually leading zero bytes. Base58 is a pure number, so a leading zero has no numeric weight and disappears unless it is deliberately re-added as a literal 1. Implementations that skip that truncate any input starting with a zero byte.
When the result has to survive case-insensitive systems or be spoken, typed or written down. DNS labels and TOTP secrets both use Base32 for exactly that reason. It is about 20 percent longer than Base64 in exchange.
Density. Four bytes become five characters rather than Base64's three becoming four, so it is roughly 7 percent smaller. PDF and PostScript use it internally. The trade-off is an alphabet containing characters that need escaping in many other contexts.
The data decoded successfully but is not text — it is probably a key, a hash or compressed data. That is reported rather than shown as substitution characters, because the latter looks like a successful decode of the wrong content.
Text is converted to bytes with UTF-8 before any encoding happens, because reading UTF-16 code units instead mangles every character outside Latin-1 — an emoji becomes two surrogate halves and round-trips to rubbish. Base32 follows RFC 4648: five bits per character from a 32-symbol alphabet, consumed in five-byte groups producing eight characters, padded with equals signs to a multiple of eight, and verified here against the RFC's own test vectors from the empty string through to foobar. Base58 uses the Bitcoin alphabet, which removes 0, O, I and l precisely because those are the pairs a person mistranscribes when copying an address by hand. It is a pure base conversion through arbitrary-precision integers, which creates a subtlety worth stating: leading zero bytes carry no numeric weight and would simply vanish, so the format re-adds each one as a literal 1 and the decoder restores them. Implementations that skip that step silently truncate any input beginning with a zero byte, which is most binary keys. Base85 packs four bytes into five characters, making it denser than Base64, and a group of four zero bytes is written as the single character z — the format's one compression trick and the part most implementations forget. Partial groups emit one more character than they had bytes. Decoding validates every character against its alphabet and names the offending one rather than producing quiet nonsense, and text that decodes to bytes which are not valid UTF-8 is reported as such instead of being replaced with substitution characters that look like a successful decode of the wrong data. Everything runs in your browser.