TonePerfect Pronunciation API
TonePerfect Pronunciation API scores how well a learner pronounced a known sentence, for language learning apps. Send a recording (up to 120 seconds) and its reference text to one REST endpoint (POST /v1/assess) and get, in one JSON response, 0-100 grades for the sentence, every word and every sound, the expected versus heard phonemes, word timestamps and a correction decision per word.
10 languages: Mandarin Chinese, English (US), French, Spanish (Latin America), German, Italian, Portuguese (Brazil), Russian, Korean and Japanese. Mandarin is scored per syllable as initial, final and tone, with tone sandhi taken into account. Built for scripted practice; it is not speech-to-text and does not score fluency.
50 free credits with no card; plans from $10/month (Starter, 2,500 credits) to $150/month (Scale, 50,000 credits). Audio is deleted after scoring and never used for training. OpenAPI 3.1 spec, Python and Node.js clients, and starter projects for Next.js, Expo and Flutter. Made by the team behind the TonePerfect pronunciation app.