Douyin Script Prompt Builder

A short-video script lives or dies in the first three seconds

When you ask a large language model — DeepSeek, Qwen (通义千问), Doubao, Kimi, ERNIE or Zhipu — to write a short-video script for Douyin, Video Account or Kuaishou, the quality of the script depends far more on how you brief it than on which clever phrase you use. A good brief names the topic or product, fixes the duration, designs the opening hook, identifies the audience, sets the voice, lays out the structure, and lists the keywords to weave in. This builder keeps that brief structured for you: fill the fields, and it joins them into a clean prompt that opens with the goal — "Write a short-video script about…" — followed by clearly headed sections, each prefixed with a Markdown-style heading the model can read at a glance. The result is the kind of brief a careful short-video producer would write by hand, only assembled in seconds and ready to paste into any model.

The single highest-leverage element is the hook. On these platforms, completion rate is decided in the first three seconds, so a vague opening quietly kills the whole clip no matter how good the rest is. A concrete hook — built around conflict, a pain point or a counter-intuitive claim — gives the model a sharp target instead of a generic "today I will teach you a method". After the hook, duration does the next bit of heavy lifting: 15 seconds can only land one point, 30 seconds allows a small build-up, and 60 seconds has room for a full arc, so fixing the length keeps the model from over-stuffing or padding. Naming the audience then keeps every line pitched correctly, whether the viewer is a first-job professional or a seasoned buyer. A good rule of thumb is to make each field concrete and specific.

Structure and the call to action turn a clip into a result

The fields people skip and regret are structure and the call to action. A clear structure — hook → body → call to action — gives the model a spine to write along, so the clip builds instead of meandering; without it, even strong lines drift. The call to action is what converts attention into action: the more specific it is — "follow for part two", "tap the cart", "drop a comment with your city" — the better it performs, while a vague "remember to support me" is almost the same as saying nothing. None of this limits the model; it focuses it. Spending one extra line on exactly what you want the viewer to do, and when, is consistently the cheapest way to lift a script from watchable to effective.

Because the output is structured plain text, the same prompt is portable across every major Chinese model and works just as well on ChatGPT, Claude or Gemini. Write it in Chinese when you want natural, idiomatic Chinese lines for a Chinese-speaking audience; the structure travels regardless of language. Ask the model to return three columns — shot, voiceover and caption — and you get something you can hand straight to filming and editing. And because the whole tool runs locally in your browser, you can iterate freely — tweak the hook, change the duration, copy again, and test — without anything you type ever leaving your device, being sent to a model, or being stored. Treat the first script as a draft: run it, see where attention drops, and tighten the matching field. Two or three rounds of that usually turn a flat clip into one that holds viewers to the end, and you keep a clean, reusable brief at the end.