Writing prompts for text-to-dialogue: an audio guide to dialogue structure

Present a "prompt structure" guide in audio form: break the input of elevenlabs/text-to-dialogue into a dialogue array, using two voices of different timbres as mutual narrator and questioner, turning "how to organize the lines of text, how to assign voices to multiple speakers, and how stability and the language field affect delivery" into a listenable tutorial. Structurally, the dialogue array provides {text, voice} line by line: one voice takes on the main narration, the other takes on follow-up questions and restatement, forming a conversational rhythm rather than one-way reading; stability takes a balanced value to balance emotional variation with voice consistency; language_code is explicitly set to Chinese to avoid misjudgments caused by automatic detection. Use cases: for developers to quickly understand the input structure before integrating the model for the first time; it can also serve as a prototype for in-product "sample audio" or documentation voiceover. Note that this pipeline outputs audio only and does not generate visuals, so this piece conveys the information of a visual-and-text guide in a purely auditory way.