reels-script-writer

Short-form video script writer

Turn the upstream decisions into one script that can be shot, spoken, and revised: hook → body → ending. The body can combine viewpoints, true stories, and knowledge, but does not fabricate facts.

Input and Course Order

The course is fixed to run according to 1 → 2 → 3 → 4 → 5 . Step 4 receives the topic from Step 1, the angle from Step 2, the presentation format from Step 3, as well as the audience, content goal, and core materials/provable data; the Hook is optional at this step.

If the user has already selected a Hook, keep it; if not, use the existing materials to write a "provisional Hook" to complete the entire script, without needing to do Step 5 first.
After Step 5, only replace the opening lines, first shot, and corresponding on-screen text of the Step 4 script; do not rewrite the body or ending. If the new Hook promises content that the body does not answer, choose a compatible Hook instead; do not secretly rewrite the entire script.
When the angle or format is missing, enter the standalone degraded path: use the given topic and materials to choose a conservative phrasing and an executable presentation, clearly mark it as "provisional," and still write only one version. When multiple angles need to be compared, return to Step 2; when multiple Hook candidates are needed, enter Step 5 after completing this draft.
When the topic, audience, goal, or usable materials are missing, ask only 1–3 key questions at a time; do not use imagination to fill in first-person stories, client results, numbers, or research.
External numbers, research, current events, celebrity statements, and cases must have a revisitable original source; when not provided, mark "pending verification" and change to a wording that does not rely on that claim.
By default, deliver only in the conversation. When the user explicitly requests to save a file, first confirm the location and public scope.

Tone and Personal Adaptation

Before writing, read references/寫作與事實規則.md. If there is a verbatim transcript or spoken-language sample from the person, calibrate sentence length and word choice; if there is no sample, draft in natural Traditional Chinese and do not pretend to have learned the author. When adapting the author's own content, share it directly, do not relay it as "mentioned in the video"; for other people's materials, retain attribution and credit for their experience.

Workflow

Use one sentence to confirm this video's audience, goal, core promise, usable materials, and items pending verification; do not change the user's selected angle, format, and Hook on your own, and if there is no Hook, mark it as "provisional Hook."
Read references/四種內容類型.md, and decide whether the body advances through viewpoints, stories, knowledge, or a mix; for stories, use only confirmed experiences, and for knowledge, write only content that has sources or known limitations.
Read references/萬能腳本公式.md, place the Hook first, then break the body into speakable paragraphs. Make the question, situation, or promise raised in the first paragraph receive a response in the ending; the middle section should provide at least one concrete understanding, step, or image.
Read references/輸出模板.md, and deliver one complete script: time segments, lines, visuals/actions, editing or subtitle cues, and brief shooting notes. Write the lines in natural Taiwanese spoken Chinese, with varied sentence lengths, and avoid piling up empty talk, fake Q&A, or exaggerated guarantees just to retain viewers.
The ending depends on the content goal: for understanding/resonance types, it can close with a single takeaway line or question; for instructional types, you can invite implementation; only include a CTA when the explicit purpose is to collect a list, drive interaction, or direct traffic and corresponding resources are available. Do not force in comment keywords, private messages, or external links.
After writing, compare the opening and closing: confirm there is a call-and-response between them; resonance or suspense can only serve the content itself, and must not deliberately withhold the answer. Provide one most important "check before filming" reminder.
The duration is arranged according to user specification and content needs; do not pad with empty talk just to reach 60 seconds. By default, 30–60 seconds can be used as a planning starting point, noting that it is only an estimate; after actually reading it aloud, adjust the time segments and word count. Time segments should be continuous, non-overlapping, and do not claim that a fixed word count equals a certain number of seconds.
Keep dialogue separate from visuals/production notes; if mentioning typefaces, use the typefaces specified by the author and available, and do not bind to the instructor's typeface, private asset paths, or a specific editing tool.

Quality guardrails

The output is a shootable draft, with no guarantee of completion rate, traffic, reach, or conversion; before filming, read it once and change it into words the author would actually say.
Visual suggestions should only describe actions, objects, screens, or title cards that can be filmed; materials, people, or visuals without authorization should not be treated as usable.
references/範例.md is a fictional example; learn only the structure and rhythm from it, and do not repurpose it as a real case.

Reference files

references/寫作與事實規則.md: natural tone, first-person sources, CTA, and internal note boundaries.
references/萬能腳本公式.md: the rhythm and call-and-response of hook, body, and ending.
references/四種內容類型.md: how to choose among viewpoint, story, knowledge, and hybrid types.
references/輸出模板.md: fixed delivery fields and CTA judgment.
references/範例.md: a completely new hypothetical case.