Personal · 2026
PronunciationTrainerAI
A speech tool that evaluates how a word or phrase was pronounced rather than whether the user knew the correct vocabulary.
Python · speech · AI
Why I built it
A lot of language tools tell you whether you produced the expected word. I was more interested in whether the individual sounds were produced correctly and where pronunciation started to diverge.
How it works
The user speaks a word or phrase, the system processes the audio, compares the spoken output against the expected pronunciation, and returns feedback around the sounds that need attention.
The project is written in Python and is designed to work across languages rather than around a single vocabulary set.
Outcome
It became an early version of an idea I later explored much more deeply in production speech systems: turning raw audio and model outputs into feedback that is specific enough for a person to act on.