Identifiers
Create a batch or inspect existing IDs. Generation is manual; decoding can update automatically.
Leave empty to sample the current millisecond once for this batch. Otherwise enter an integer from 0 to 281474976710655; ISO dates and seconds are not inferred.
Processing stays in this browser. Inputs are not sent to a server.
On: draw the first 80-bit tail, then increment it for each ID. Off: draw each tail independently. Every click starts a new batch; there is no ordering promise between batches.
1–1,000 IDs; decode up to 1,000 physical lines and 100,000 UTF-16 code units. Exports include the entire result.
Choose settings, then run.
A ULID generator helps when you need sortable test records or want to understand an ID in a log. The useful question is often whether its embedded time and its position in a batch agree, rather than whether the string merely has 26 characters.
Choose a fixed millisecond to reproduce a time prefix without fixing the random tail. When inspecting existing values, the report keeps the original line number even after sorting, so an unexpected position can be traced back to the input.
The ULID generator separates the ten-character timestamp from the sixteen-character random tail. UTC and the original integer milliseconds appear together, so a seconds-versus-milliseconds mistake does not hide behind a formatted date.
Sorting produces a report without moving input lines. Each decoded record retains sourceLine, including repeated values; you can compare a changed position with the exact value originally supplied.
The generation example loads a fixed timestamp but does not generate IDs. Empty time means the current instant at execution; a value such as 1700000000 is interpreted literally as milliseconds.
Monotonic mode increments the random tail inside this batch. Turn it off to see how independent tails at the same millisecond can sort differently from their generation order.
In Decode, Example loads deliberately shuffled values. Switch Result order to ASCII ascending and compare sourceLine. Auto decoding waits 350 ms; disable it for the Decode button or Ctrl/Cmd+Enter.
Generate 1–1,000 standard uppercase identifiers. Fixed time fixes only the first ten characters; the remaining eighty bits come from secure random bytes, followed by increments when monotonic mode is on.
The ULID decoder reports the ULID timestamp as Unix milliseconds and UTC ISO time, along with both encoded parts and the random tail in hexadecimal. Nothing in that tail identifies a user or machine.
A malformed character, length error or overflow reports its physical line and column. The whole decode fails, so copying cannot silently discard invalid entries while presenting a smaller successful batch.
Generation exports one ID per line as TXT, or a JSON batch with its time and monotonic flag. Decoding exports all field records, including rows beyond the initial fifty-row preview.
Canonical uppercase strings sort by embedded millisecond using ASCII order. Within one millisecond, independent random tails need not follow creation order. Enable monotonic mode for a strictly increasing sequence within this one batch; distributed clocks still require separate coordination.
A ULID has 26 Crockford Base32 characters and a 48-bit millisecond prefix. A UUID has a versioned layout and commonly uses 36-character hexadecimal text. Equal storage width does not make ULID text valid UUID text or imply a particular UUID version.
The ULID generator draws one random tail and increments it for later IDs in the same batch. Every new generation action starts independently. It does not remember a previous batch or correct a backwards clock across clicks.
For independently random IDs sharing a timestamp, the random space is 2^80 and collisions are possible. Monotonic tails are distinct inside one successful batch. Neither mechanism guarantees global uniqueness across independent generators; enforce uniqueness where your data requires it.
The first ten characters encode the Unix millisecond time in 48 bits; the first character uses only three significant bits. This timestamp can be chosen by the generator, so recovering it is not proof of an actual insertion or event time.
Have a question about an ID?
Use the ULID generator with a fixed test time, then decode a shuffled sample to compare timestamp order and original line positions before exporting.