A sentence to read
With the recording and the sentence, OpenPronounce returns the four scores, every word with its sounds and timings, and the words that went wrong.
From the command line
$ openpronounce developer.wav "Hello, I am a good designer" Score : 36.2/100 accuracy 36.6, fluency 55.0, prosodic 45.9 Warning : not said: good (scores unreliable) Transcription: HELLO I AM A DEVELOPER Heard phones : /h ə l oʊ aɪ æ m ə d ɪ v ɛ l ə p ɚ/ Mispronounced: - good: not said - designer: expected /dɪzaɪnɚ/, heard /vɛləpɚ/ (confidence 89%)
--json prints the whole result, --lang fr changes language, --fast runs the sound recognizer alone.
From Python
from openpronounce import load_audio, assess result = assess(load_audio("developer.wav"), "Hello, I am a good designer", lang="en")
load_audio reads wav, mp3, flac, ogg or webm and returns a 16 kHz mono waveform. If you already have one, pass the numpy array directly.
What it reports
- Mispronounced words, in
differences.errors: the sounds expected and heard, and the probability that the raters would have rejected the word. A word is reported above 0.35. - Words not said, in the same list with
missing: true. 83 in 100 are found. - Speech in excess, in
differences.extra, with the sounds heard and where they fall. - Another sentence altogether, in
differences.alignment: the recording is checked against the text before anything else.
Faster, lighter
| Option | Skips | Use it when |
|---|---|---|
--fast | Transcription, synthetic reference, prosody curves | You need a score in under a second |
acoustic=False | The synthetic native reading | No network at all, or you do not need formula_score |
prosody=False | Pitch and energy curves | You do not draw them |
transcription=False | The word transcription | You only need the scores and the sounds |