Pronunciation Analyzer

What does “say it more clearly” actually mean? A speech-analysis application that relates a recording to a reference script and makes the feedback more specific.

Recognition → alignment → phoneme feedback

What I worked on

I combined Whisper ASR with forced alignment and dynamic time warping to map spoken audio onto reference phonemes. Pitch analysis through Praat/Parselmouth adds a prosody view alongside phoneme-level accuracy and fluency scores.

The result is an interactive HTML report: click a word to replay its audio, inspect phoneme tooltips, and compare sessions stored in SQLite. The application was deployed on a Linux VPS with Flask, gunicorn, nginx, and systemd, using a scripted deployment.

The cover is an illustration, not a learner recording or a measured assessment result.

Tools and methods Python · Whisper · Forced alignment · DTW · Praat / Parselmouth · Flask