Mang is a generator tool inspired by the donjon 5e Tools and Kate Monk's Onomastikon.
It started originally as a project for quickly generating names for D&D NPCs, and became a tool I used for writing and game development. Other generation capabilities have been added over time to fit new needs that popped up.
The name generation tool uses Markov chains (hence the name Markov Name Generator) to build words similar to, but not quite, names from the real world.
Other generation tools use a mix of whatever random generation approach seems feasible for the tool at hand. Feel free to play or make suggestions.
Changelog
2025-12-26
Mang v1 (Strict)
- Names no longer jump randomly when stuck; Mang now re-paths to create more natural phonetic flows.
- Introduced temperature to occasionally pick less common letter combinations to avoid repetitive or plain results.
- Added back checks to block unpronounceable clusters (like triple consonants or vowels).
- Automatically filters out exact copies of names from the input to ensure output is original.
- Increased length variance and transition logic to produce a wider range of length in output.
2025-12-25
Mang v1 (Strict):
- Generated words now start and end more naturally based on training data patterns
- Improved word length calculation uses average length from samples with natural variation
- Better quality control: validates vowel-consonant balance and avoids excessive repeated characters
- Removes diacritics from input (é→e, ñ→n, ü→u, etc.)
- Strips non-alphabetic characters like dashes, apostrophes, and spaces from training data
- Generated words are more pronounceable and resemble sample patterns
- Multiple generation attempts ensure higher quality output