Skip to content

Mondegreen (Soramimi) Finder

Reads a line in one language by that language’s spelling rules, nativises it onto another language’s sound system, and ranks the phrases in that language that sound closest — rule-based, offline, with the alignment shown.

🔒 Runs entirely in your browserphoneticsipalanguagewordplaykorean

The text

Detection reads the script and the letters and nothing else. Pick one yourself if it gets it wrong.

Finds what the line sounds like to somebody whose ears are tuned to that language.

What this does and does not do

The text is read with that language’s spelling rules into IPA, then rewritten to fit the sound system of the language doing the listening — Korean puts a ㅡ inside an initial cluster and reduces the finals to seven, Japanese puts a vowel after every consonant. That string of sounds is then matched phoneme by phoneme against the target language’s word list.

The score is how close the sounds are, not how funny the result is. It is a distance over place, manner and voicing for consonants and over height, backness and rounding for vowels, so /p/ and /b/ are close and /p/ and /i/ are far apart. A 90 can easily be less funny than a 60.

No model, no pronunciation dictionary, no network. Every rule is written in the source and each language says where its rules came from. A language it cannot do is named and refused rather than fudged: Vietnamese, Thai, Arabic, Hebrew, Hindi, Portuguese, Chinese (hanzi), Japanese (kanji). If you know the pronunciation, use the direct-IPA mode.

The heavy work runs in a worker so typing never stutters.

Feed it from

Next

These accept what this tool produces, so the result can go straight in.