A vowel-space and formant-analysis project that grew out of a Master's thesis, "Towards Accounting for L2 Accent: The Case of Turkish Vowel Space" — a study of how native English speakers produce the Turkish vowel system compared to native Turkish speakers, and what that says about the line between phonological knowledge (the abstract system) and phonetic performance (the motor habits of actually producing it).
Formerly called AudioForge.
Live: ayter.com/timbre Author: Can Ayter
Ten speakers (five native Turkish, five L1-English learners) produced 37 Turkish words anchoring the language's eight vowel phonemes. Formants were measured in Praat at the stable middle portion of each vowel. The learners' vowels came out more centralized than their native English inventory, but still missed the Turkish targets in a consistent way — particularly a more closed /a/ and a more peripheral /i/ — suggesting the gap is mostly in motor execution transferred from English, not in understanding Turkish's vowel harmony system itself.
A written presentation of that research, plus a static IPA cardinal-vowel chart (Jones, 1917) you can export as SVG or PNG, and a reference table spanning 41 languages and roughly 300 vowel phonemes across five language families (Turkic, Uralic, Mongolic, Indo-European, Sino-Tibetan), drawn from published acoustic studies.
Being direct about scope: the page's own copy describes a fuller interactive toolkit — live formant extraction from a microphone, spectrograms, MFCC extraction, an accuracy score against target vowels — as the vision for the tool. As deployed right now, that live analyzer isn't implemented; there's no microphone capture or LPC computation running in the current index.html. What's genuinely live is the thesis writeup and the static reference chart. Closing that gap is the natural next step, not a hidden bug.
timbre/
├── index.html the page: research writeup, cardinal vowel chart, reference table
├── _headers hosting header configuration
└── legacy-streamlit-prototype/
an earlier Python/Streamlit prototype (audio_processor.py,
classifier.py, feature_extractor.py) built before the project
moved to a single static page. Kept for history, not maintained.
Can Ayter — ayter.com
MIT License.