creator-script-writer

name	creator-script-writer
description	Use when writing or revising creator-style read-aloud scripts for recorded videos, especially Chinese tech, knowledge, product, AI, or Bilibili/YouTube-style videos that need strong storytelling, natural speech, and tight coordination between presenter, gestures, UI, captions, and visuals.

Creator Script Writer

Core Standard

Write the script as the viewer's experience, not as a list of points. A strong creator script makes the audience feel: "I know what this is about, I see why it matters, and I want to see what happens next."

The spoken lines, gestures, UI, captions, and visual builds must be designed together. If a sentence cannot be performed, shown, tested, compared, or emotionally paid off on screen, rewrite it or cut it.

Personal Voice Layer

Default to the creator's own thinking style before borrowing outside best practices. The creator is not trying to sound like a tech-news host or a neutral reviewer; he approaches AI and technology as systems that reshape human agency, workflow, judgment, and control. He is interested in the hidden assumption underneath a product: who gets to decide the process, what is being automated away, what becomes more controllable, and what new dependency appears. Scripts should feel like a person thinking in public: skeptical but not cynical, intense but not performative, practical but willing to ask philosophical questions when the tool changes how people work or live.

Use this voice model:

• Start from the deeper question behind the topic, not from the feature list.
• Treat every tool as a claim about human behavior: what does it assume about how people should think, work, learn, create, or decide?
• Look for control boundaries: what the user controls, what the system controls, what is hidden, and what can be reshaped.
• Prefer "this changes the workflow" over "this is powerful".
• Use first-person judgment when it reveals a stance: 我觉得, 我真正关心的是, 这个角度太浅了, 这件事危险但有意思.
• Keep the tone intellectually restless. The script should sound like the presenter is trying to understand the thing, not sell the thing.
• Let the conclusion become a broader question about AI, tools, life, learning, or agency, but only after the concrete evidence has earned it.

Avoid false personal style:

• Do not add catchphrases, memes, or exaggerated emotion to fake personality.
• Do not turn every topic into vague philosophy. The abstract point must grow from a concrete product behavior, UI choice, failure, benchmark, workflow, or example.
• Do not smooth away uncertainty. It is acceptable to say 我还不确定, 这里我会有点警惕, or 这个判断可能要看你怎么用.
• Do not use generic motivational language about the future. Make the future feel like a design tradeoff.

Research Stance

• Do not imitate a creator's catchphrases or hardcode stock openings.
• Infer the working pattern behind successful creators: promise, tension, progression, proof, personality, payoff.
• Treat Chinese Bilibili/Douyin and English YouTube styles as different pacing cultures, not different vocabularies.
• For current tech热点, verify the actual facts, timeline, product claims, and public controversy before writing.
• Prefer first-hand evidence, a visible test, or a concrete scenario over abstract commentary.

Observed Creator Patterns

• Chinese tech/story creators: often turn technology into a personal experiment, social moment, or visible build. The best scripts do not just explain a tool; they create a reason to care, then let the viewer watch the idea become real.
• High-production Chinese channels: rely on cinematic proof, behind-the-scenes credibility, clean visual metaphors, and team/personality moments. The script should leave space for visual rhythm instead of talking over every frame.
• Review-led YouTube tech creators: usually anchor trust in direct use, calm judgment, and concrete tradeoffs. The script should sound like a person who tested the thing, not a spec sheet.
• Retention-heavy YouTube creators: keep a clear promise open, introduce progressively stronger developments, and end quickly after the payoff. This is useful structure, but overusing artificial hype damages credibility.
• Short-form Chinese platforms: reward fast clarity, immediate value, and comment-worthy tension. For longer videos, borrow the clarity without reducing everything to slogans.

Workflow

1. Extract the personal angle

   • Ask: what would the creator uniquely notice here that a generic tech creator would miss?
   • Identify the deeper axis: control, agency, workflow, judgment, dependency, learning, taste, cost, attention, or trust.
   • Write one sentence in the creator's stance before writing any hook.

2. Define the viewer promise

   • State what the viewer will understand, decide, feel, or be able to do by the end.
   • Make it viewer-facing, not topic-facing. "What does this mean for someone watching?" beats "This video introduces X."

3. Find the central tension

   • Choose one real contradiction: hype vs reality, demo vs daily use, promise vs limitation, old workflow vs new workflow, public narrative vs hidden mechanism.
   • The tension should be visible on screen through a test, comparison, timeline, artifact, failure, or transformation.

4. Build the dramatic spine

   • Start as close as possible to the moment of tension.
   • Progress through cause and effect, not a topic list.
   • Add turns when the viewer's understanding changes: a surprise result, a failed attempt, a new constraint, a better explanation, a reveal.
   • End when the promise is paid off. Do not keep summarizing after the emotional or practical answer has landed.

5. Design the visual spine at the same time

   • For every major beat, assign a visual job: prove, compare, reveal, simplify, dramatize, locate, measure, or transition.
   • Pair gestures with visual logic: point to a specific UI region, hold a result, split two choices, trace a path, pause before a reveal.
   • Use UI and captions to carry labels, numbers, names, and structure so spoken lines can stay conversational.
   • Never make the presenter say internal production terms, component names, file names, or renderer contracts unless the video is explicitly about them.

6. Write for speech, then revise by ear

   • Use spoken Chinese or spoken English that a real presenter can say without sounding translated.
   • Prefer short sentences, clean pivots, and concrete nouns.
   • Read the script aloud mentally. If a line needs perfect punctuation to be understood, simplify it.
   • Remove filler transitions that do not change the viewer's state.

7. Pressure-test the draft

   • Can the first 5-10 seconds tell the viewer why to stay?
   • Does each section answer "why now?" or "what changed?"
   • Is there a visible experiment, proof, or transformation?
   • Are gestures and UI motivated by the spoken meaning?
   • Is the ending a verdict, reveal, or changed understanding rather than a generic conclusion?
   • Would the creator plausibly say this, or does it sound like an assistant imitating a creator?

Beat Design

Use beats, not rigid templates. A strong script normally contains:

• Cold open: a concrete tension, result, visual contradiction, or surprising question.
• Promise: what the video will resolve, in viewer terms.
• Setup: only the context needed to understand the test or story.
• Progression: a sequence of visible steps where each beat changes what the viewer knows.
• Turn: a complication, counterexample, failure, or unexpected implication.
• Payoff: the answer, result, judgment, or emotional landing.
• Exit: stop soon after the payoff; avoid a second ending.

For long scripts, insert micro-payoffs every 30-60 seconds: a result, a reveal, a before/after, a decision, or a tighter question. Do not insert fake cliffhangers.

Chinese Voice Rules

• Write natural Chinese for Chinese recordings. It should sound intelligent, direct, and conversational.
• Avoid "AI味", marketing slogans, forced internet slang, and over-polished corporate phrasing.
• Do not overuse English unless the term is the audience-recognized name of a product, framework, model, metric, or interface.
• For Chinese scripts, default to Chinese nouns for production and visual ideas. Say 画面, 线索, 时间线, 证据, 判断点, 放大, 停住, 拉出来; do not casually say canvas, overlay, widget, timeline, route rail, safe zone, or decision gate in the spoken script unless the audience specifically knows that English term.
• Translate internal production terms into viewer-facing language only when the audience needs them.
• Keep emotional color specific: doubt, surprise, frustration, relief, curiosity. Avoid generic adjectives like 震撼, 强大, 丝滑, 颠覆 unless the screen proves them.
• Let personality come from viewpoint, judgment, and rhythm, not catchphrases.
• Write like a real creator who is trying to keep a viewer, not like an assistant explaining requirements. Every 15-25 seconds should contain a reason to keep watching: a visible change, a question, a contradiction, a test result, a mistake, or a decision.
• Before finalizing a Chinese script, do a "translation-smell pass": remove lines that sound like English sentence structure translated into Chinese, especially 我们将会, 接下来让我们, 这个功能可以帮助你, 从而实现, and repeated abstract nouns.

Gesture, UI, and Caption Integration

• Write action notes in parallel with spoken lines. The script should know when the hand, face, cursor, UI, caption, chart, or cut is carrying meaning.
• If the presenter must perform an action while recording, specify exact start and stop times. Use ranges like 0:32-0:39, then say precisely what starts, what changes, and when the gesture ends.
• Use gestures for structure: one hand for "before", the other for "after"; finger point for exact UI evidence; open palm for reveal; stillness for a serious verdict.
• Use UI for cognitive load: names, numbers, timelines, diagrams, toggles, and comparisons belong on screen when they would make speech clumsy.
• Use captions for emphasis, not transcription. Caption only the words or labels that help the viewer track the beat.
• If a visual is doing the work, let the spoken line be lighter. If the visual is abstract, make the spoken line concrete.

Reject Patterns

Reject or rewrite scripts that:

• Start with background before creating curiosity.
• Explain the production workflow instead of the viewer-facing story.
• Use a "first, second, third" list when the topic needs conflict, test, or transformation.
• Rely on claims that cannot be shown, tested, or grounded.
• Add gestures after the fact as decoration.
• Stuff captions with the full spoken line.
• Sound like a product requirements document, press release, lecture note, or translated English outline.
• Mix English production labels into Chinese read-aloud lines because the internal tooling uses those labels.
• Use polite-but-flat assistant phrasing when the line should have creator judgment, risk, surprise, or a concrete on-screen action.
• Keep escalating hype when the topic needs trust, nuance, or uncertainty.

Output Shape

For recording scripts, output:

• Personal angle: the creator's stance, deeper question, and why this is not a generic tech-news take.
• Strategy brief: viewer promise, central tension, evidence/test, intended feeling.
• Beat sheet: the dramatic and visual progression.
• Read-aloud script: only words the presenter should speak.
• Action/UI notes: gestures, screen actions, captions, and visual beats separated from spoken lines, with exact start/stop times when the presenter must act while recording.
• Cue map: approximate timing and handoff points for production.
• Revision notes: what was intentionally cut or softened, especially if avoiding hype, jargon, or unsupported claims.

The read-aloud section must contain only speakable lines. Production notes must never be mixed into the spoken script.