About

Raw data values - timestamps, byte counts, large integers - are unambiguous to machines but hostile to human cognition. A value like 1048576 requires mental parsing; the phrase "one million forty-eight thousand five hundred seventy-six" does not. Misreading a single digit in a financial figure or a date field can cascade into contractual errors, incorrect invoices, or compliance violations. This tool implements deterministic verbalization algorithms for 10 distinct data types: cardinal and ordinal numbers (up to 10¹⁵), ISO dates, relative timestamps, millisecond durations, Oxford-comma lists, binary and SI file sizes, booleans, percentages, and English pluralization with an irregular-word dictionary.

Each function returns a plain English string with no markup. The number-to-words engine uses recursive scale-group decomposition through trillions. The pluralizer covers 80+ irregular English nouns. Relative time calculations reference Date.now() and bucket deltas into the most natural human unit. This tool approximates natural speech patterns. It does not handle locale-specific grammar or non-English languages. Pro tip: when verbalizing currency, convert to the minor unit (cents) first to avoid floating-point rounding artifacts.

Formulas

The number-to-words algorithm decomposes an integer n into scale groups using integer division and modulo. For each scale s in the ordered set {trillion, billion, million, thousand}:

chunk = floor(n ÷ s)

n = n − chunk × s

Each chunk (0 - 999) is recursively verbalized into hundreds, tens, and ones using lookup tables. Decimals are handled by splitting at the decimal point and verbalizing each digit of the fractional part individually, joined by the word "point".

Relative time computes the signed delta Δt:

Δt = t_now − t_input

The delta is bucketed through thresholds: 60 s → minutes, 3600 s → hours, 86400 s → days, 2592000 s → months, 31536000 s → years. The sign of Δt determines "ago" vs. "in".

File size conversion divides bytes by the appropriate base:

value = bytesbase^e

Where base = 1024 for binary (KiB, MiB, GiB) or 1000 for SI (KB, MB, GB), and e is the largest exponent such that value ≥ 1.

The ordinal suffix is determined by the last two digits d of n: if d ∈ {11, 12, 13} → "th"; else last digit 1 → "st", 2 → "nd", 3 → "rd", otherwise "th".

Where n = input number, s = scale value, chunk = group value (0 - 999), t_now = current timestamp, t_input = input timestamp, Δt = time difference in seconds, bytes = input byte count, base = 1024 or 1000, e = scale exponent, d = last two digits for ordinal suffix.

Reference Data

Mode	Input Example	Output Example	Range / Limits
Number to Words	1450	one thousand four hundred fifty	−999 trillion to 999 trillion
Ordinal	42	42nd	Any non-negative integer
Date to Words	2024-03-15	March 15th, 2024	Any valid ISO 8601 date
Relative Time	2024-01-01T00:00:00Z	5 months ago	Any valid date/timestamp
Duration	90061000 ms	1 day, 1 hour, 1 minute, 1 second	0 to 10¹⁵ ms
List	apples, oranges, bananas	apples, oranges, and bananas	Any comma-separated items
File Size	1073741824 bytes	1.00 GiB	0 to 10¹⁵ bytes
Boolean	TRUE	yes	TRUE / FALSE / 1 / 0
Percentage	42.5	forty-two point five percent	Any numeric value
Pluralize	child, 3	3 children	English nouns, count ≥ 0
Number (negative)	−250	negative two hundred fifty	Handles sign prefix
Number (decimal)	3.14	three point one four	Arbitrary decimal digits
Duration (small)	450 ms	450 milliseconds	Sub-second precision
File Size (SI)	1000000 bytes	1.00 MB	SI (base-10) mode
Ordinal (edge)	111	111th	Handles 11th/12th/13th correctly

Frequently Asked Questions

English teens are irregular: "eleven" through "nineteen" do not follow the tens-plus-ones pattern of numbers like "twenty-one." The algorithm uses a dedicated lookup table for values 11-19. When the remainder after extracting hundreds is between 11 and 19, the engine indexes directly into this table instead of decomposing into a tens word plus a ones word. This prevents outputs like "ten-one" for 11.

The converter caps at ±999,999,999,999,999 (just under 10¹⁵). Values beyond this range return a clear error message rather than producing incorrect output. JavaScript's Number.MAX_SAFE_INTEGER is approximately 9 × 10¹⁵, so precision loss in the input itself is the primary risk. For financial or scientific values exceeding this range, convert the number to a string representation and process externally.

The default is binary (KiB, MiB, GiB) using base 1024, which matches how operating systems report storage. A toggle switches to SI (KB, MB, GB) using base 1000, which matches how drive manufacturers label capacity. The discrepancy means a "500 GB" drive shows as approximately 465 GiB. Choosing the wrong base for a capacity report can create a perceived 7-10% error in reported values.

The engine maintains a dictionary of 80+ irregular plural forms (child → children, mouse → mice, datum → data, person → people, etc.). Before applying suffix rules (-s, -es, -ies), it checks the irregular dictionary for an exact match. Uncountable nouns (sheep, fish, species) return the same form regardless of count. Words not in the dictionary fall back to standard English suffix rules: consonant+y → -ies, sibilants → -es, otherwise -s.

The converter accepts ISO 8601 strings (2024-03-15, 2024-03-15T14:30:00Z) and Unix timestamps in milliseconds. Dates without a timezone suffix are parsed in the browser's local timezone per the ECMAScript spec. To avoid ambiguity, append "Z" for UTC or include an explicit offset like "+05:30". The relative calculation uses Date.now() as the reference point, so results update if you re-trigger conversion.

The converter truncates input to whole milliseconds. Sub-millisecond values (microseconds, nanoseconds) are not standard in JavaScript's Date API and are outside the scope of human-readable duration text. For values under 1000 ms, the output reads as milliseconds directly (e.g., '450 milliseconds'). For zero, it returns "zero milliseconds". Negative durations are treated as their absolute value with a "negative" prefix.