README

One short recording per sound of American English — 24 consonants, 11 vowels, 5 diphthongs — to play next to an IPA symbol in pronunciation lessons, and the example words the same speaker says right after each sound. Every file comes from a source listed below, under that source's license; this README, credits.json and the ID3 tags of each file carry the credit.

Listen to them all: https://vocamax.github.io/ipa-audios/ (GitHub does not play audio inside a README).

The files are cut from recordings of real speakers. Text-to-speech cannot say a sound on its own, and the IPA charts with audio on Wikimedia use cardinal vowels, not American ones.

Sounds

40 of 40 sounds have a file.

Consonants (24/24)

Vowels (11/11)

Diphthongs (5/5)

¹ License as labeled by the uploader; the original source of the recording is not confirmed yet.

Example words

119 words for 40 sounds, said by the same speaker right after the sound. They are the words of the source video, not the examples of sounds.json.

Sources

Learn IPA - American English - Diphthongs

• Source: https://www.youtube.com/watch?v=65lGhp4IKug (sources/youtube-65lGhp4IKug.json)
• Author: Elise Linh learn English
• License: CC BY 3.0 — as labeled by the uploader, not confirmed
• Notes: Six diphthongs (aʊ oʊ ɔɪ aɪ eɪ ju): each is said alone when its symbol appears, then in three words; the "Diphthong Review" (1:06–1:17, on a title card) says all six again in one take, in near-digital silence. The five clips come from the review; [ju] is not one of the 40 sounds. Same speaker and series as sources/youtube-vzaY6tkqhjg.json; the channel's consonant video has another speaker — confirm that the uploader owns the recording before relying on the CC BY label.
• Example words: 15 (segments in credits.json)
• Retrieved: 2026-11-12

Learn IPA - American English - Consonants

• Source: https://www.youtube.com/watch?v=YG1-7V2rlZQ (sources/youtube-YG1-7V2rlZQ.json)
• Author: Elise Linh learn English
• License: CC BY 3.0 — as labeled by the uploader, not confirmed
• Notes: The 24 consonants, in the order d t v f b p g k θ ð s z ʃ ʒ tʃ dʒ h ɹ j m n ŋ l w: each is said alone one to three times while only its symbol is on screen, then in three words; there is no review at the end. Every clip is the first time the sound is said alone. Another speaker than the vowel and diphthong videos of the same channel, which looks like a compilation of recordings from elsewhere — confirm that the uploader owns the recording before relying on the CC BY label. /θ/, /h/, /f/ and /p/ are only 5–15 dB above the noise floor of the recording and needed a large gain; /f/ has less padding after it because "staff" follows at once.
• Example words: 72 (segments in credits.json)
• Retrieved: 2026-11-12

Learn IPA - American English - Vowels

• Source: https://www.youtube.com/watch?v=vzaY6tkqhjg (sources/youtube-vzaY6tkqhjg.json)
• Author: Elise Linh learn English
• License: CC BY 3.0 — as labeled by the uploader, not confirmed
• Notes: The 11 vowels: each is said alone when its symbol appears, then in three words; the "Vowel Review" (1:52–2:13) says all 11 again in one take. A Vietnamese title card comes before the footage and the channel is small — confirm that the uploader owns the recording before relying on the CC BY label. /æ/ comes from the introduction: the review take measured closer to /ɑ/.
• Example words: 32 (segments in credits.json)
• Retrieved: 2026-11-12

Changes made to the source material

Each file is cut to one sound or word (about 0.1 s of silence before it, 0.2 s after), loudness-normalized to a mean of −20 dBFS, peak-limited, faded in over 20 ms and out over 60 ms, and encoded as mono MP3 at 96 kbps. Nothing else is changed. The exact segment of every file is in the tables above and in credits.json.

Using the files

• sounds/<key>.mp3 — keys are in sounds.json (p, th, iy, er, ow…).
• words/<key>/<word>.mp3 — example words of a sound, listed under words.<key> in credits.json (the word as written on screen, e.g. March; the file name is in lower case).
• credits.json — for every file: IPA symbol or word, SHA-1, source title and link (with the time of the segment), author, license and whether the source has been verified. Show the credit wherever the sounds are used, or link to this README: https://github.com/vocamax/ipa-audios.
• With Composer: composer require vocamax/ipa-audios, files under vendor/vocamax/ipa-audios/. Pin a tag: a new tag can replace a file.

Adding a source

Needs PHP 8.2+ with GD, ffmpeg, ffprobe and yt-dlp.

1. php bin/scan <youtube url> — prints the license, every speech segment with suggested cut points and the caption words around it, and writes .cache/youtube-<id>/sheet.png (a frame from the middle of each segment) and manifest.draft.json. Videos without a Creative Commons license on YouTube cannot be used without permission.
2. Copy the draft to sources/youtube-<id>.json and add one clip per sound: sound (key from sounds.json), from and to (seconds, padding included) and label (what the video shows). Label each segment by the symbol on screen in sheet.png, not by counting. One file per sound across all sources: move a sound by deleting its clip from the old manifest. Example words go in words: sound, word (as written on screen), from, to, label.
3. php bin/build — downloads into .cache/, cuts, writes sounds/, words/, credits.json and the tables of this README. A warning about a large gain means the clip is mostly silence (or a very weak sound such as /θ/, /h/).
4. Listen: php -S 127.0.0.1:8124 -t . and open http://localhost:8124/.cache/preview.html.
5. php bin/build --check before committing. On every push to main, .github/workflows/pages.yml runs the check and publishes the listening page (php bin/site _site) to GitHub Pages.

Sources other than YouTube: "kind": "url" (direct link to an audio file, e.g. Wikimedia Commons) or "kind": "file" with "path" (a recording committed to this repository).

verified stays false until someone has confirmed that the uploader owns the recording and chose its license; a false source is flagged in the tables and in credits.json.

License

Each file keeps the license of its source, shown in the tables above. CC BY: give the credit (title, author, link, license, changes) where the sound is used. CC BY-SA: the same, and the cut files stay under CC BY-SA.