Back to Details

say-to-her

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Say To Her

讲给她听

Transform any source material into a single dramatic, entertaining conversational snippet — complete with stage directions — that a programmer can confidently deliver to his girlfriend.
将任意素材转化为一段富有戏剧性和趣味性的对话片段——附带表达指导——让程序员可以自信地讲给女友听。

Core Principles

核心原则

  • ONE nugget only. Scan the source, find the single most interesting or surprising point, discard everything else. Never summarize.
  • Dramatize ruthlessly. Exaggerate, add flair, make it theatrical. Entertainment value over accuracy.
  • Stage directions are MANDATORY. Every output must include inline delivery annotations — pauses, voice cues, gestures, expressions. No exceptions.
  • Storytelling, not lecturing. Sound like an excited boyfriend sharing something wild, never like a teacher or presenter.
  • No jargon leakage. Translate all technical terms into casual language or wrap them in playful explanation.
  • No moralizing. Suppress all "this raises important questions about..." instincts. Pure entertainment.
  • Output length: 100-250 words, speakable in 30-90 seconds.
  • Output language: match the user's prompt language. If they write in Chinese, the script is in Chinese. If English, English.
  • Default audience: smart non-technical person who is curious but has zero context on the topic.
  • 只保留一个亮点：浏览素材，找到最有趣或最令人惊讶的点，其余内容全部舍弃。绝对不要做总结。
  • 大胆戏剧化处理：夸张表达、增添亮点，使其更具戏剧性。娱乐性优先于准确性。
  • 必须包含表达指导：所有输出都必须内嵌表达标注——停顿、语调提示、手势、表情。无一例外。
  • 用叙事而非说教：要像兴奋的男友分享新奇趣事，绝对不能像老师或演讲者那样说话。
  • 杜绝术语泄露：将所有技术术语转化为日常用语，或用有趣的方式解释。
  • 不做道德说教：克制住“这引发了关于……的重要问题”这类想法。只专注于娱乐。
  • 输出长度：100-250词，口语化表达时长为30-90秒。
  • 输出语言：匹配用户提示的语言。如果用户用中文提问，脚本就用中文；如果是英文，就用英文。
  • 默认受众：聪明的非技术人群，他们充满好奇心，但对相关主题毫无背景知识。

Transformation Workflow

转化流程

Transforming source material into a script follows these steps:
  1. Accept source material — pasted text, topic description, article excerpt, any format. Do NOT fetch URLs; work only with provided text.
  2. Scan for the ONE most interesting nugget — look for what's surprising, counterintuitive, viscerally impressive, or just plain weird. Ignore everything else.
  3. Discard the rest — ruthlessly. If the source has 10 interesting facts, pick 1. Commit to it.
  4. Dramatize with storytelling structure:
    • Hook — grab attention: "So you know how...?" / "Okay, get this—" / "Wait, I read the craziest thing today"
    • Tension — build curiosity, set up the reveal, make her lean in
    • Reveal — deliver the punchline, the mind-blowing fact, the "wait WHAT?" moment
    • Closer — playful tag-out line, callback, or "right??" to land the ending
  5. Annotate with delivery stage directions throughout the script — not just at the start. See 
    references/tone-guide.md
    for the full annotation catalog. Aim for roughly 1 annotation per 2-3 sentences.
  6. Format as a ready-to-use script using the output template below.
将素材转化为脚本需遵循以下步骤：
  1. 接收素材——粘贴的文本、主题描述、文章节选，任意格式均可。不要获取URL链接；仅基于提供的文本进行处理。
  2. 寻找最有趣的一个亮点——关注那些令人惊讶、违反直觉、冲击力强或非常奇特的内容。忽略其他所有信息。
  3. 舍弃其余内容——坚决执行。如果素材里有10个有趣的事实，只选一个。专注于此。
  4. 用叙事结构进行戏剧化处理：
    • 钩子——吸引注意力：“你知道吗……？” / “哎，跟你说个事儿——” / “哇，我今天看到个超疯狂的事情”
    • 制造悬念——引发好奇心，为揭秘做铺垫，让她产生兴趣
    • 揭秘——抛出核心亮点，带来“哇？真的假的？”的时刻
    • 收尾——用俏皮的结束语、呼应前文的内容或“对吧？”来完美收尾
  5. 在脚本中全程嵌入表达指导——不只是在开头。完整的标注目录请参考
    references/tone-guide.md
    。目标是每2-3句话就加入一个标注。
  6. 按照下方的输出模板格式化为可直接使用的脚本。

Output Template

输出模板

Always format output as follows:
undefined
请始终按照以下格式输出：
undefined

Your Script

你的脚本

Topic: [one-line summary of the nugget] Vibe: [emotional flavor, e.g. "mind-blown revelation", "cute fun fact", "conspiracy-theory dramatic"] Best delivered: [context suggestion, e.g. "over dinner", "random text", "while cooking together"]
[The script with inline (delivery annotations) woven throughout]
Pro tip: [one practical delivery suggestion, e.g. "Pause after the big number — let her do the math in her head" or "If she laughs at the hook, ride that energy into the reveal"]
undefined
主题：[亮点的一句话总结] 氛围：[情感风格，例如“令人震惊的发现”、“可爱的冷知识”、“阴谋论式的戏剧性”] 最佳场景：[场景建议，例如“晚餐时”、“随机发消息”、“一起做饭时”]
[内嵌（表达指导）的脚本内容]
小贴士：[一个实用的表达建议，例如“在说出关键数字后停顿一下——让她自己在心里算一算”或“如果她对钩子部分笑了，顺着这个情绪进入揭秘环节”]
undefined

What NOT to Produce

禁忌内容

Avoid these anti-patterns at all costs:
  • The TED Talk — informative but boring, no personality, sounds like a presentation. Missing: hooks, stage directions, casual language.
  • The Wikipedia Entry — accurate but dry, tries to cover everything instead of cherry-picking one nugget. Missing: storytelling arc, dramatization.
  • The Lecture — talks down to her, jargon-heavy, explains too much background. Missing: respect for the audience, entertainment value.
  • The Robot — no stage directions, flat delivery cues. The words might be good but there's no guidance on HOW to say them.
See 
references/examples.md
for full side-by-side comparisons of bad vs. good transformations.
务必避免以下错误做法：
  • TED演讲式——信息丰富但枯燥，没有个性，听起来像正式演讲。缺失：钩子、表达指导、日常用语。
  • 维基百科条目式——准确但乏味，试图涵盖所有内容而非精选一个亮点。缺失：叙事结构、戏剧化处理。
  • 说教式——以居高临下的语气说话，充满术语，解释过多背景知识。缺失：对受众的尊重、娱乐性。
  • 机器人式——没有表达指导，语气平淡。内容可能不错，但没有关于如何表达的指导。
完整的反面案例与正面案例对比请参考
references/examples.md
。

Edge Cases

边缘情况处理

  • Source is already simple/casual — still dramatize. Don't parrot; add storytelling structure and stage directions.
  • Source is very long — fine. Just pick ONE nugget from it. Length of input has no bearing on output.
  • Source is in a different language than user's prompt — translate and transform in one step. Output in the user's language.
  • Source has no obvious interesting nugget — find the least boring bit and dramatize it hard. If truly nothing works, tell the user and suggest different material.
  • Same source provided again — pick a DIFFERENT nugget this time. Never repeat.
  • Sensitive or controversial topic — transform it, but prepend: 
    (read the room before telling this one)
    at the top of the script.
  • 素材本身已简单/口语化——仍需进行戏剧化处理。不要直接照搬；要添加叙事结构和表达指导。
  • 素材篇幅很长——没问题。只需从中挑选一个亮点即可。输入长度不影响输出。
  • 素材语言与用户提示语言不同——一步完成翻译和转化。输出语言与用户提示一致。
  • 素材没有明显的有趣亮点——找到最不无聊的部分并进行强戏剧化处理。如果确实毫无可用内容，请告知用户并建议更换素材。
  • 重复提供同一素材——这次要挑选不同的亮点。绝对不要重复。
  • 敏感或有争议的主题——可以进行转化，但需在脚本顶部添加：
    (讲述前先观察对方的反应)
    。

References

参考资料

Load these reference files when generating output:
  • references/examples.md
     — Complete source-to-script transformation examples (4 good, 2 anti-patterns) demonstrating the exact tone, structure, and annotation density expected. Load this on first use or when calibrating output quality.
  • references/tone-guide.md
     — Full delivery annotation catalog organized by category (pauses, voice, gestures, expressions, pacing, engagement hooks) with usage guidance and combination rules. Load this when selecting and placing stage directions.
生成输出时请参考以下文件：
  • references/examples.md
    ——完整的素材转脚本示例（4个正面案例，2个反面案例），展示了预期的语气、结构和标注密度。首次使用或校准输出质量时请参考此文件。
  • references/tone-guide.md
    ——完整的表达指导目录，按类别划分（停顿、语调、手势、表情、节奏、互动钩子），并附使用说明和组合规则。选择和添加表达指导时请参考此文件。