Friendly reading aid · ARPABET · broad IPA
Phonetic Spelling Generator
See how a covered word can be read aloud without hiding the evidence: every retained pronunciation variant, its friendly respelling, the original ARPABET sequence, a derived IPA view, and the stress behind it.
Enter one word. If the retained CMUdict source covers it, the generator shows every retained variant in three forms: an informal pronunciation spelling, source ARPABET, and a broad IPA conversion. Capital letters in the friendly spelling mark primary stress. The result is a reading aid from one primarily North American source, not audio or a universal dialect verdict.
Look up one word
How do you pronounce it?
Use a standard dictionary spelling. The lookup never invents a pronunciation when this release has no source record.
Method, in the open
How the phonetic spelling is generated
The useful question is not just “what symbols appeared?” but which facts came from the pronunciation source and which parts are display transformations. This generator keeps that chain visible.
-
01
Resolve one spelling
The input is trimmed, lowercased, and checked as one ASCII-letter word. Internal apostrophes are removed for the compiled lookup. A missing record stays missing; the tool does not guess from spelling.
-
02
Keep every retained variant
The loaded pronunciation sidecar returns its primary ARPABET sequence and every alternative retained for the same spelling, in source order. Each row is processed independently.
-
03
Make a friendly reading aid
A fixed ARPABET-to-letter table turns vowel and consonant symbols into familiar chunks. Consonants are grouped with the following vowel; the chunk containing primary stress is capitalised.
-
04
Map ARPABET to broad IPA
Each stress-stripped ARPABET symbol maps through the project's fixed IPA table. No acoustic model, text-to-speech engine, or hidden language model changes the result.
Interpretation guide
Three formats, three different jobs
“Phonetic spelling” can mean an informal pronunciation key or a formal phonetic transcription. Showing both without labels would create false precision, so every result names the role of each form.
| Format | Best for | What it does not claim |
|---|---|---|
| Friendly respelling | A quick English-language reading cue. Hyphens separate deterministic chunks and capitals point to primary stress. | It is not a standard alphabet, exact mouth instruction, or reliable dictionary syllable division. |
| ARPABET | Auditing the actual retained CMUdict phoneme sequence. Digits on vowel symbols preserve unstressed, primary, and secondary stress. | It is source notation rather than familiar spelling, and the record is not evidence for every English dialect. |
| Broad IPA conversion | Reading the source phonemes in a widely used phonetic alphabet and comparing sound identities across ordinary spelling differences. | This conversion does not add aspiration, vowel length, allophones, syllable boundaries, or IPA stress marks. |
Stress evidence
What 0, 1, and 2 mean
CMUdict writes a digit on every vowel phoneme. Reading those digits from left to right gives the displayed stress pattern and, because each marked vowel is one nucleus, the derived syllable count.
0- Unstressed. The vowel nucleus does not carry lexical stress in this variant.
1- Primary stress. The strongest lexical beat; its friendly chunk is uppercase.
2- Secondary stress. A weaker lexical stress, shown in evidence but not capitalised by the friendly renderer.
Limits before confidence
What this generator cannot hear
A pronunciation record is useful evidence, but it is not a recording of every speaker. Treat the output as a documented starting point for saying a covered word, not as a judgement that another established pronunciation is wrong.
Source coverage
The retained CMUdict data is primarily North American English and incomplete. Personal and place names, new coinages, specialist terms, compounds, inflected forms, loanwords, and regional variants can be absent.
Broad symbols
The IPA row is a phoneme-by-phoneme conversion, not a narrow transcription made from audio. It leaves out predictable physical detail such as aspiration, reduction choices, linking, and speaker-specific vowel quality.
Friendly ambiguity
Ordinary letters are familiar precisely because they are flexible. Chunks such as uh, th, or oo cannot encode every contrast as precisely as a phonetic alphabet; use the IPA and ARPABET rows to check.
No audio or speech diagnosis
The tool does not play a model speaker, listen through a microphone, score an accent, or provide speech therapy. It transforms one pinned text source into labelled visual aids.
Pronunciation results use the loaded lexical release shown in the report and site footer. See data sources and licences for the release boundary, transformation notes, and source limitations.
Questions, answered
Phonetic spelling and transcription FAQ
What is a phonetic spelling generator?
It turns a retained pronunciation into an easier-to-scan spelling such as huh-LOH. This tool also shows the source ARPABET and a broad IPA conversion so the friendly aid is never mistaken for a formal transcription.
Can this tell me how to pronounce any word?
No. It reports only spellings covered by the retained CMUdict artifact. Names, new coinages, compounds, loanwords, inflections, and dialect forms are incomplete, and no spelling-based guess is substituted for missing evidence.
Why are there several pronunciations?
CMUdict can retain more than one pronunciation for a spelling. The variants may reflect different stress, vowel choices, or uses, but they are not an exhaustive catalogue of every accent or meaning.
Is the IPA a complete phonetic transcription?
No. It is a deterministic broad symbol mapping from ARPABET. It omits fine phonetic detail, and this converter presents stress separately rather than inserting IPA stress marks or asserted syllable boundaries.
Does the generator play audio or hear my voice?
No. It neither synthesizes audio nor records speech. The outputs are visual reading aids derived from a pinned pronunciation source.
Is the word I enter stored?
No. It travels in a bounded POST body, is looked up in memory, and is returned in a no-store response. It is not put into a URL, application log message, or saved draft.