H3 电影感双镜有声短片（纯文本）H3 cinematic two-shot scene with native audio (text-only)

纯文本生成有声短片：第一镜建立场景与人物，第二镜在指定时间点转折。提示词按 MiniMax H3 的三段式结构组织——画面、环境声、配乐分开写，避免模型把音乐当成环境音处理。

Text-only two-shot cinematic scene with native audio: shot one establishes, shot two turns on a named timecode.