interview.get_questions
ChatGPTUse this when someone wants common English job-interview questions to answer aloud. Returns general questions, not a company-specific or specialist question bank.
practice.list_scenes
ChatGPTList Speaking Coach's live practice scenes with their ids, partner, goal and length, to choose a sceneId for practice.scene.
practice.open
ChatGPTOpen Speaking Coach's practice menu: live role-play conversations, Judge your speech, Hear your potential, presentation rehearsal and quick voice reviews. Use when the user wants to practice speaking but has not picked a specific activity. Opens controls only.
practice.scene
ChatGPTOpen a live spoken role-play with an AI partner who plays the other person: a job interview (optionally with the user's CV or job post attached in the widget), asking for a raise, disagreeing in a meeting, setting a boundary, a rude customer, tough feedback, bad news, a first date, networking, a pitch, or a situation the user describes. Afterwards the user gets a report with scores quoting their words, how they sounded (pace, filler words, freezes), a better way to say one line, and one focused retry. Pass sceneId from practice.list_scenes when one fits; otherwise pass customDescription and customPartner. Scenes last 2–6 minutes. Opens controls only; the user consents and starts the call in the widget.
presentation.rehearse
ChatGPTRehearse a presentation over the user's own slides (PDF or PowerPoint, text only), with the slide text on screen. Afterwards the audience asks 4–6 questions grounded in the deck and the talk; the user answers each aloud and gets one adjustment per answer, plus their pace and filler words. Up to 9 minutes. Opens controls only; the user uploads slides in the widget.
review.compare
ChatGPTCompare measured pace, filler words and pauses in two completed recordings of the same task. Changes do not imply that faster speech or fewer pauses are better.
review.get_result
ChatGPTValidate and return a completed voice review from the recorder. Requires the signed result returned after audio processing; does not accept raw audio or an arbitrary transcript.
review.start
ChatGPTUse this for a quick review of one short spoken answer (1–3 minutes): measured pace, filler words, pauses and one quoted change. To judge or score a whole talk, toast, wedding speech, pitch or presentation, use speech.judge instead; to rehearse with slides, presentation.rehearse; to role-play a conversation, practice.scene. Opens an English-language recorder with optional PDF, DOCX, PPTX, text or Markdown attachment for context. File parsing reads text only; it does not evaluate slide visuals. Do not use for text-only editing, translation, accent or pronunciation coaching. The recording is limited to 1–3 minutes by mode. When opening practice, omit documentProof; users upload their file inside the recorder. Never invent a document proof or supply a filename as one.
speech.judge
ChatGPTUse this whenever the user wants a talk, toast, wedding speech, class presentation, pitch or speech judged, scored or rated. Scores a talk the user gives on their own (a toast, wedding speech, class presentation, pitch or answer) out of 100, with notes on opening, structure, clarity and close, the one thing to keep, the one fix, and measured pace, filler words and freezes. Up to 9 minutes, recorded or uploaded in the widget. Opens controls only.
speech.potential
ChatGPTThe user speaks freely for 20–90 seconds, then hears the same message at its best (fillers removed, main point first, calmer pace) read in a temporary copy of their own voice, next to their original. The voice copy is deleted straight after and needs its own consent in the widget. Only for the user's own voice. Do not call it when someone asks to copy, imitate or analyze another person's voice or recording; decline instead. Opens controls only.