Output Transforms

The output transform system modifies generated text after sentence construction. Text is tokenized into words, punctuation, and whitespace. Protected tokens (acronyms, numbers, URLs) are marked, then each transform in the pipeline processes the token stream in order.

How it works

1. Tokenize – The generated text is split into typed tokens (word, punctuation, whitespace)
2. Protect – Certain tokens are marked as protected (acronyms, numbers, code tokens, URLs)
3. Transform – Each transform in the pipeline processes the token stream sequentially
4. Render – Tokens are joined back into a string

Available transforms

ID	Description
pigLatin	Classic Pig Latin: consonant clusters move to end + “ay”
ubbiDubbi	Ubbi Dubbi language game: inserts “ub” before each vowel sound
leet	Leetspeak: substitutes characters (a->4, e->3, o->0, etc.)
uwu	Cute speak: r/l -> w substitution, nya insertion, cute suffixes
pirate	Pirate speak: word and phrase substitutions
redact	Redaction: masks words with bracketed blocks
emoji	Emoji: replaces words with emoji equivalents
mockCase	Mock case: rAnDoM CaSe aLtErNaTiOn
reverseWords	Reverse: reverses word order within sentences
bizJargon	Business jargon: applies business-speak patterns

See Built-in Transforms for detailed descriptions and examples.

Quick example

In code

const result = generator.sentence({
  outputTransforms: {
    enabled: true,
    pipeline: [{ id: 'pigLatin' }],
  },
});

From the CLI

malarky sentence --seed 42 --transform pigLatin
# "Enerallygay, ethay angechay alledcay."

All transforms are deterministic (same seed = same output) and safe to chain. See Chaining & Configuration for pipeline ordering and protection rules.

---

Table of contents

• Built-in Transforms
• Chaining & Configuration
• Custom Transforms