Random Text Generator — random words, sentences, names & prompts

Five pieces work together: a curated common-English corpus, a set of grammatical sentence templates, a pair of international and Sri Lankan name banks, a bank of original writing prompts, and a small deterministic pseudo-random number generator that ties them all together. Every step is a pure function — no network calls, no local storage, no analytics on the generated text itself.

1. Word corpus. About 553 common English tokens drawn from the intersection of the Oxford 3000 and the General Service List (West, 1953; modern ELT successor lists). All lowercase, no proper nouns, no profanity. Word mode picks n tokens uniformly at random and applies your chosen letter case — lower, Title, or UPPER.
2. Sentence construction. Sentences are built from 18 grammatical English templates with placeholders filled from sub-banks of 121 concrete nouns, 113 present-tense verbs, 49 adjectives, and 33 adverbs. About 14% of period-ended templates have their terminator swapped for a ? or ! to add rhythm. Paragraphs string 3–6 sentences together and join with a single space; multiple paragraphs are joined with exactly one blank line (\n\n).
3. Names. The tool ships two name corpora — an international bank (91 first names + 95 surnames) and a Sri Lankan bank (62Sinhala & Tamil first names + 46 surnames). Sri Lankan names are transliterated to plain Latin script so the output renders the same on every browser. Toggle the Sri Lankan switch to mix or unmix. The tool tracks seen full names inside a single run and retries on collisions, so duplicates are rare at counts under 100.
4. Writing prompts. About 60 original prompts hand-written for this tool. Each is one complete English sentence ending with a period or question mark. Prompts are sampled without replacement within a single run — request 60 or more and you receive the entire bank in shuffled order. Topics span memory, character portrait, place, ritual, and craft warm-ups; nothing is AI-generated.
5. Deterministic PRNG. The seed feeds a Mulberry32 32-bit PRNG (Ettinger, public domain). Calling generate(...) twice with the same seed and options returns byte-identical output. The seed is shown next to the action row so you can copy it down and reproduce a particular run later — useful for snapshot tests, design-system docs, and reproducible bug reports.

Verification: on every render the page calls verifyDeterministic() which generates the same options with the same seed twice and compares the strings character by character. When both runs match, the “Verified · deterministic” badge lights up in the card header. Counts are also clamped to per-mode limits (1,000 words / 200 sentences / 50 paragraphs / 200 names / 50 prompts) so very large inputs never lock the tab.

• Oxford Learner's Dictionaries — Oxford 3000 & 5000 word lists
• Wikipedia — General Service List (West, 1953) and successor ELT lists
• US Census Bureau — frequently occurring surnames (2010)
• US Social Security Administration — popular baby names
• UK Office for National Statistics — baby names in England & Wales
• Mulberry32 PRNG — Tommy Ettinger's reference implementation

The corpora and generation logic on this page were last verified on 2026-05-11. Word and name banks are reviewed annually or when an upstream source updates. Writing prompts are curated locally and added quarterly.