Skip to content

ReferenceFormatting & media

Snowflakes

How Discord ids are built, with a decoder for timestamps.

9 entries Snowflakes

Bit layout 4

Field	Bits	Size	Extract	Notes
Timestamp	63 to 22	42 bits	`(snowflake >> 22) + 1420070400000`	Milliseconds since Discord Epoch, the first second of 2015 or 1420070400000.
Internal worker ID	21 to 17	5 bits	`(snowflake & 0x3E0000) >> 17`
Internal process ID	16 to 12	5 bits	`(snowflake & 0x1F000) >> 12`
Increment	11 to 0	12 bits	`snowflake & 0xFFF`	For every ID that is generated on that process, this number is incremented

Handy values 5

What	Value	Notes
Discord epoch	`1420070400000`	First second of 2015 in Unix milliseconds.
Timestamp from id	`(id >> 22) + 1420070400000`	Milliseconds since Unix epoch. Use BigInt, ids exceed 2^53.
Id from timestamp	`(ms - 1420070400000) << 22`	Lowest possible id for a moment; handy for `before`/`after` pagination.
Serialization	`string`	Ids are always sent as strings in JSON to avoid precision loss.
Default avatar index	`(user_id >> 22) % 6`	For migrated usernames. Legacy discriminator users use `discriminator % 5`.