A useful tutor responds to how a learner spoke, not only what the transcript says. Chivox turns each attempt into structured evidence so the tutor can choose one useful cue, keep the conversation warm and invite a focused retry.
The learner speaks naturally, assessment runs between turns, and the tutor returns with one correction that is tied to actual evidence—not a detached results screen.
Scores become useful only when they change the next prompt, example, hint or practice activity. The tutor should choose a priority, explain it simply and keep momentum.
Assessment is the stable source of speech evidence. Your prompts, lessons, interface voice and retry logic can evolve independently as you learn what helps learners improve.
A learner has completed a conversation about travel. The tutor does not interrupt the whole exchange with a dashboard. It notices one repeated sound in a word that matters, gives credit for what worked and invites a small retry.
“I sink the train leaves at three.”
The sentence is smooth enough to keep the conversation moving.
The first sound is repeated and changes the meaning of the word.
Correct one sound before introducing a new prompt or drill.
“Nice, that was very smooth. One small sound: for think, let a little air pass over the tip of your tongue—th-ink. Try just the word once, then say the sentence again.”
Implementation note. Chivox identifies the speech evidence; your teaching approach decides the cue, encouragement, retry length and when to move on.
The tutor does not need to repeat every score. It can select the highest-value issue, explain it in the learner’s context and generate a retry that keeps the conversation moving.
{
"assessment": {
"overall": 82,
"priority": { "word": "think", "phoneme": "θ", "score": 54 },
"audio_quality": 96
},
"tutor_action": {
"type": "focused_retry",
"cue": "Keep your tongue lightly between your teeth for /θ/.",
"next_prompt": "Try: I think they are ready."
}
}Move from a natural learner turn to one teaching priority and a focused retry without losing conversational context.
Transcription answers what the system heard. Pronunciation assessment adds evidence about how the utterance was produced, including segmental and fluency detail.
Yes. The orchestration can stay consistent while routing each activity to language-specific assessment and feedback logic.
Tell us what you’re building. We’ll reply within one business day with pilot credits, pricing, or a deployment plan — whichever you need first.