Choose ONE scenario and design the voice flow. Return JSON {"scenario":"...","voice_steps":"...","ai_role":"..."}, where scenario is one of: language_learning (language learning), email_draft (letter draft), meeting_notes (meeting notes); voice_steps — steps of voice work; ai_role - what role you assign to the assistant in advance.
The voice is especially strong in three everyday scenarios, and each has its own mechanics. Language learning: Conversational AI works as an infinitely patient interlocutor - speak with mistakes, and ask it to gently correct and repeat the correct version out loud so you can hear the sound. Technique: ask them to repeat your phrase in a corrected form, then you learn by ear, and not by writing. Draft a letter by voice: do not dictate the finished text - state your intention in abstracts (“write to a colleague: reschedule the meeting for Friday, politely, briefly”), and give the polishing to the model. It's faster and more natural. Notes on the meeting: dictate the results immediately after the conversation, while still in memory, then arrange them into a structure. General insight: in voice scenarios, the division of labor wins - the person gives meaning and intention with the voice, the model gives form and purity. The second trick: for any scenario, agree in advance with the assistant on the role (“you are my teacher”, “you are my editor”) - the given role sharply increases hitting the desired tone. A common mistake is to use your voice as a typewriter and dictate the final text letter by letter; its strength is the speed of the draft and the dialogue.
Unlock access to submit solutions for instant AI review.