Short drama script · turn content into shots

Disclaimer: All content shown on this page is generated by AI models and is for reference only; it does not constitute any guarantee of results. Actual output and Credits consumption are subject to "Usage Details" in the console.

You have a piece of material or a content draft and want to turn it into a voiceover short video you can shoot right away. The hard part isn't "writing it smoothly"—it's three things that easily go wrong: the duration must be precise, the lines must be readable end-to-end, and the numbers must not be fabricated. This is especially true for knowledge content—a single wrong figure sinks the whole piece. Call chain: qwen3.8-max (deep thinking) · A "dangers of sitting too long" explainer draft · Ask it to adapt the content into a 60-second voiceover short video script, output shot by shot in JSON (each shot includes visual / narration / seconds) · Require that every number in the narration comes only from the original draft and may not be fabricated. The output in one line: a hook "Office workers who sit too much—don't scroll past, this concerns you" + 11 shots, each with visual/narration/seconds, totaling exactly 60 seconds—see the full storyboard below.

Step: Lock down "no fabricated numbers"

Send the original draft along with the following instruction to qwen3.8-max:

You are a short video scriptwriter. Please adapt the explainer draft below into a 60-second voiceover short video script for a knowledge-focused account.
Output strictly in JSON:
{
  "title": "Video title (click-worthy, not exaggerated, not clickbait)",
  "hook": "Hook copy for the first 3 seconds (one sentence)",
  "shots": [ {"no": 1, "visual": "Visual description", "narration": "Voiceover line", "seconds": 5}, ... ],
  "total_seconds": 60, "bgm": "Soundtrack mood", "cta": "One closing engagement line"
}
Hard requirements:
1. The seconds of all shots must add up to exactly 60
2. Every number and fact in the narration must come from the original draft below—no improvising, no fabricating data; do not write numbers that aren't in the original draft into the lines
3. Lines must be conversational and readable as-is; match the word count of each line to its seconds (normal speaking rate is about 4-5 characters per second)
4. Keep at least 3 key data points from the original draft
5. The hook should capture the target audience but not sell anxiety to the point of falsehood
Original draft: [Paste your content draft / material here]

View the result · full storyboard script (raw output from qwen3.8-max)

Title: Self-rescue for office workers who sit too much: 9.3 hours a day—what now?
Hook: Office workers who sit too much—don't scroll past, this concerns you.

Shot  Duration  Visual / Voiceover line
---------------------------------------------------------
1    3s    Desk close-up, looking into the camera
            "Office workers who sit too much—don't scroll past, this concerns you."
2    5s    Programmer/designer/finance icons, progress bar pulled to 9.3 hours
            "Programmers, designers, finance staff sit an average of 9.3 hours continuously every day."
3    5s    Pie chart pops up 41%, clock fast-forwards past 2 hours
            "Among them, 41% don't get up until they've sat for over 2 hours at a time."
4    7s    Lipoprotein lipase activity drops 90% animation
            "After 90 minutes of sitting, lipoprotein lipase activity drops by about 90%—the same diet makes fat storage easier."
5    5s    Leg blood vessel diagram, blood flow speed bar drops about 50%
            "After sitting for 3 hours straight, leg blood flow drops by about 50%."
6    6s    Lumbar spine pressure comparison chart
            "Sitting lumbar pressure is about 1.4x standing; slouching brings it to 1.9x."
7    8s    Research data chart: 800,000 people, 8 hours, 59%
            "A study of 800,000 people: those who sit over 8 hours a day and barely move have a 59%
             higher all-cause mortality risk than those who sit less than 4 hours."
8    6s    Brisk walking, cycling footage, subtitle 60-75 minutes
            "60 to 75 minutes of moderate-intensity activity a day can nearly offset this risk."
9    4s    Timer 20 minutes, sit 8 / stand 2 demo
            "Try this: out of every 20 minutes, sit for 8 and stand for 2."
10   4s    After a full hour of sitting, get up for water, subtitle at least 5 minutes
            "After every full hour of sitting, get up and move for at least 5 minutes."
11   7s    Host on camera summary, "stand up" animation popping in comments
            "The harm isn't sitting—it's staying still for long stretches. Move a little every so often. Drop 'stand up' in the comments."
---------------------------------------------------------
Total 60 seconds  ·  Soundtrack: upbeat, rational explainer feel  ·  CTA: Drop "stand up" in the comments

Notes

Consumption: This playbook involves text generation + deep thinking (qwen3.8-max). This kind of "multiple hard constraints" task requires enormous thinking: the body here is only 16 KB, but the thinking ran over 30,000 words—it iteratively adjusted the seconds and word counts of each shot while cross-checking the source of every number. The output isn't long, but don't estimate cost by word count.
Why JSON output: Once each shot's seconds and narration are split into fields, duration and speaking rate can be verified precisely with a few lines of code, and the output can be directly fed to editing / voiceover tools. It's far more controllable than a plain-text storyboard.
Speaking-rate threshold reference: The comfort zone for Chinese voiceover is about 4-5 characters per second; anything over 6.5 characters per second is basically unreadable. You can write this into the prompt as a self-check.
Number sourcing: The key constraint is the line "numbers may only come from the original draft; don't write numbers that aren't there"—this must be preserved for knowledge content with data, or the model will readily invent a plausible-looking number that has no source.

​Step: Lock down "no fabricated numbers"

​Notes

Step: Lock down "no fabricated numbers"

Notes