Trình tạo ULID

(Tài liệu bằng tiếng Anh)

What it does

Generates ULIDs: 26-character identifiers that pair a millisecond timestamp with 80 bits of randomness. The tool fills itself in the moment it loads and re-rolls on every option change, so there is always a result on screen. Set Count from 1 to 1,000. A single value renders in large type; a batch renders as a list. Copy takes the whole batch. Everything runs locally with crypto.getRandomValues.

How to use it

1. Set Count (1 to 1,000). Results regenerate as you change it.
2. Read the result: one value in large type, two or more as a list.
3. Press Copy to take the whole batch.
4. Paste any

   ULIDULIDA 128-bit identifier like a UUID, but timestamp-prefixed and Crockford-base32 encoded — so it sorts by creation time and is shorter to read.

   into Decode a

   ULIDULIDA 128-bit identifier like a UUID, but timestamp-prefixed and Crockford-base32 encoded — so it sorts by creation time and is shorter to read.

   to inspect it: the tool splits time part from randomness and shows the timestamp in milliseconds and ISO-8601 UTC. Invalid input (wrong length, non-Crockford character, or a time part past the 48-bit ceiling) shows the exact reason.

Examples

A

ULIDULIDA 128-bit identifier like a UUID, but timestamp-prefixed and Crockford-base32 encoded — so it sorts by creation time and is shorter to read.

is 26 characters. The split is the whole point:

01ARZ3NDEKTSV4RRFFQ69G5FAV
└──┬──┘└───────┬────────┘
timestamp   randomness
(10 chars)  (16 chars)

• 01ARZ3NDEKTSV4RRFFQ69G5FAV - the spec’s canonical example
• 01ARZ3NDEK... sorts before 01ARZ3NDF... because 10 characters of timestamp lead the string
• Pasting 01ARZ3NDEKTSV4RRFFQ69G5FAV into the decoder reads the timestamp 2016-07-30T23:54:10.259Z — the spec’s canonical example was minted on 2016-07-30

How ULIDs work

A

ULIDULIDA 128-bit identifier like a UUID, but timestamp-prefixed and Crockford-base32 encoded — so it sorts by creation time and is shorter to read.

is 128 bits, cut in two:

Part	Bits	Characters	Holds
Timestamp	48	first 10	milliseconds since the Unix epoch
Randomness	80	last 16	random bits

Both halves encode in Crockford base32: 32 symbols, 5 bits per character. The alphabet is 0-9 then A-Z minus I, L, O, U. That makes ULIDs case-insensitive and removes the letters people misread (I as 1, O as 0). 26 characters times 5 bits is 130, so the top 2 bits sit unused.

The timestamp is what you are buying. It occupies the most significant end of the value, so string order equals creation order: sort a column of ULIDs as plain text and you get insertion order for free. Compare with UUIDv4:

Property	ULIDULIDA 128-bit identifier like a UUID, but timestamp-prefixed and Crockford-base32 encoded — so it sorts by creation time and is shorter to read.	UUIDv4
Characters	26	36 (with 4 hyphens)
Sorts by creation time	yes, lexicographic	no - fully random order
Random bits	80	122

UUIDv4 is fine when identifiers only need to be unique. When they also need to sort - database keys, event logs, anything you page through - the timestamp prefix turns an index scan into a range read. The 48-bit millisecond clock reaches the year 10889, so it will not overflow in practice.

Same millisecond, different ULIDs: the 80 random bits keep same-instant IDs distinct with a collision chance so small it is ignored at normal volumes.

Good to know

• Sortable, not ordered. ULIDs sort by millisecond; two IDs inside one millisecond have no guaranteed order.
• Case-insensitive. 01arz3ndektsv4rrffq69g5fav parses as the same value as the uppercase form.
• Private: generated locally in your browser. Nothing is sent anywhere.
• Related tools:

  UUIDUUIDA 128-bit identifier generated to be practically unique without coordination, most commonly in version-4 (fully random) form.

  Generator, Token Generator.