SuperAPI AI Directory

podcast-generator

Writing & research

View repository

Updated 10-15 22:14

About this skill

Podcast Script Generation Guidelines

Script Format (Important!)

This format must be used:角色|情感|文本

晓晓|cheerful|欢迎来到《历史不装》,我是晓晓。
云扬|calm|我是云扬。
晓晓|chat|今天咱们聊聊楚汉争霸。云扬你先说说,很多人觉得刘邦就是人多欺负人少,你怎么看?
云扬|serious|这话听着就不对。你知道吗,这场战争其实是17打1。

Emotion Tags (Core!)

Tag Meaning Typical Scenario Typical Phrasing
cheerful Happy, cheerful Agreement, interesting discovery, success "Haha", "Great", "Exactly", "Oh!"
chat Casual chat Ordinary conversation, transition "Today let's talk about", "And then?"
calm Calm, steady Explaining, stating facts "Simply put", "Actually", "The principle is"
serious Serious, earnest Emphasizing key points, analysis "The key is", "The core lies in", "Did you know"
gentle Gentle, soft voice Comforting, empathy "It's okay", "Take your time"
fearful Surprised, worried Unexpected information, shock "Really?", "That's too...", "No way"
sad Sad, down Describing problems, pain points "How tragic", "Too hard", "No way around it"
angry Angry, dissatisfied Criticism, complaining "That's too much", "In your dreams", "Absolutely ridiculous"
disgruntled Dissatisfied, complaining Mild complaint "Domineering is domineering, but...", "Ruthless enough"

Character Settings

  • Xiaoxiao: Curious questioner, representing the audience's perspective

    • Mostly uses:cheerful、chat、fearful
    • Occasionally:sad(resonating with pain points),disgruntled(complaining)
  • Yunyang: Professional answerer, explains in plain language

    • Mostly uses:calm、serious、chat、cheerful
    • Occasionally:angry(emphasis),sad(describing a miserable situation)

Dialogue Style

Must achieve

  1. Colloquial: Like friends chatting, not reading a script
  2. Mainly short sentences: Each sentence no more than 30 characters
  3. Continuous expression: The same character speaks 2-4 sentences in a row before switching (important!)
  4. Complete viewpoint: Each turn should express a complete thought; do not switch speakers after just one sentence
  5. Life analogies: Complex concepts must be paired with analogies

Dialogue Rhythm Rules (Core!)

  • ❌ Wrong: A says 1 sentence → B says 1 sentence → A says 1 sentence (too fragmented)
  • ✅ Correct: A speaks 2-3 complete sentences in a row → B responds and expands with 2-3 sentences → A continues

Example Comparison:

❌ Wrong (too fragmented):

晓晓|chat|项羽分封了吗?
云扬|calm|分了。
晓晓|chat|分了多少?
云扬|calm|18个王。

✅ Correct (coherent):

晓晓|chat|项羽火烧咸阳之后干了啥?分封诸侯了吗?
云扬|calm|对,分封诸侯。一共分了18个王。不过分得太随意了,你猜刘邦分到哪儿了?
晓晓|fearful|关中?他功劳最大嘛。
云扬|angry|想得美!巴蜀,秦朝流放犯人的地方。关中被一分为三,全给了秦国降将。刘邦当场差点吐血,拎着剑要去找项羽拼命。

Xiaoxiao's Expression Patterns

  • Questions: "What does this mean?", "Why does this happen?"
  • Reactions: "Oh!", "Wow, so that's how it is", "Interesting!"
  • Follow-up: "So how exactly do you do it?", "Can you give an example?"
  • Surprise: "Really?", "That's amazing", "No way..."

Yunyang's expression patterns

  • Opening: "To be honest...", "Frankly speaking...", "Truth be told..."
  • Explanation: "Simply put, it's...", "You can think of it as..."
  • Analogy: "It's like...", "Just like...", "It's similar to..."
  • Emphasis: "The key is...", "The most important thing is...", "The core lies in..."
  • Example: "For instance...", "For example...", "Think about it..."

Transition and interaction words that must be used

Transition words (enhance dialogue fluency)

  • Light transition: however, but, though, only
  • Strong transition: then again, conversely, from another angle
  • Progression: moreover, more importantly, the key is
  • Contrast: by comparison, on the contrary, quite the opposite

Response words (enhance interactivity)

  • Agreement: yes yes yes, exactly, indeed, that's right, you're right
  • Surprise: for real, no way, that's outrageous, damn
  • Understanding: got it, I see, so that's how it is, no wonder
  • Follow-up: and then, go on, be specific, like what

Connecting words (make the dialogue coherent)

  • Sequential: so, therefore, that way, thus
  • Example: for instance, take... for example, look
  • Summary: put simply, in short, in one sentence
  • Introduction: speaking of this, mentioning..., talking about...

Interaction cues (close the distance)

  • Soliciting: don't you think, right, what do you think, you know what I mean
  • Resonance: we all know, we've all been through it, who hasn't
  • Rhetorical question: guess what, believe it or not, do you know why

Do not use

  • ❌ "first, second, finally"
  • ❌ "it is worth noting", "it should be pointed out"
  • ❌ em dash (replace with comma or period)
  • ❌ classical Chinese words like "the said", "this", "its"
  • ❌ more than 3 consecutive sentences without interaction
  • ❌ any single sentence over 40 characters
  • ❌ academic and official tone

Complete example (optimized version)

【访谈式播客-带情感】楚汉争霸:刘邦凭什么赢?

晓晓|cheerful|欢迎来到《历史不装》,我是晓晓。
云扬|calm|我是云扬。
晓晓|chat|今天咱们聊聊楚汉争霸。云扬你先说说,很多人觉得刘邦就是人多欺负人少,你怎么看?
云扬|serious|这话听着就不对。你知道吗,这场战争其实是17打1。
晓晓|fearful|17打1?不是刘邦和项羽两个人吗?
云扬|chat|表面上是,但你得把彭越、英布、陈平、张良、萧何、韩信这帮人全算上。少了哪一个,刘邦都得完蛋。
晓晓|cheerful|哦!所以项羽是在跟整个天下对抗?
云扬|serious|对对对,可以这么说。但就是这样一场看似一边倒的战争,打得那叫一个精彩。

【第一章:项羽埋雷】
晓晓|chat|那咱们从项羽分封说起吧。项羽火烧咸阳之后干了啥?
云扬|calm|分封诸侯啊。一共分了18个王。
晓晓|cheerful|听起来挺霸气的。
云扬|disgruntled|霸气是霸气,不过分得太随意了。你猜刘邦分到哪儿了?
晓晓|gentle|关中?他功劳最大嘛。
云扬|angry|想得美。巴蜀!秦朝流放犯人的地方。关中被一分为三,全给了秦国降将。
晓晓|fearful|这也太欺负人了吧?
云扬|serious|可不是嘛。刘邦当场差点吐血,拎着剑要去找项羽拼命。还好萧何拉住他:苟着总比死了强,对吧?
晓晓|disgruntled|所以项羽这是在给自己埋雷?
云扬|cheerful|没错!这些雷很快就炸了。

Improvements:

  • ✅ Add response words: "yes yes yes", "isn't that so", "exactly"
  • ✅ Add transition words: "however", "but"
  • ✅ Add connecting words: "then", "so"
  • ✅ Add interaction cues: "right", "guess"
  • ✅ Make the dialogue feel more like a real chat, not a Q&A

Principles for emotion annotation

  1. Annotate every sentence: don't be lazy, think about the speaker's emotional state for every sentence
  2. Match the context: decide the emotion based on what is said, don't just label randomly
  3. Natural transition: emotional changes should be reasonable, don't jump too much
  4. chat is the default: use chat for ordinary dialogue; use others only when there is obvious emotion

Voice configuration

Reference audio location

  • Female voice:workspace/voice_female.wav(real-person recording preferred)
  • Male voice:workspace/voice_male.wav(real-person recording preferred)

IndexTTS2 parameters

  • emo_weight: emotion intensity (0-1, default 0.65)
  • temperature: randomness (0.1-2.0, default 0.8)

How to use

Generate script

请根据以下内容生成双人访谈播客脚本:
[粘贴文章内容]

要求:
1. 使用 角色|情感|文本 格式
2. 遵循SKILL.md中的对话规范和情感标签
3. 晓晓提问,云扬解答
4. 口语化,有互动感
5. 每段不超过40字

Generate audio

python skill.py workspace/脚本.txt --tts-engine indextts2 -o 播客.mp3