Text-to-video prompt generator that links multiple shots through a Storyboard to generate narrative-coherent professional video prompts.
Core Constraints
- Total word count: entire Storyboard ≤ 2000 words (hard limit)
- Default total duration: 15 seconds (preferred when unspecified)
- Default shot count: 7 (recommended 6-8, prefer 7)
- Single shot duration (15-second fast cuts): 1-3 seconds
- Single shot duration (conventional narrative): 4-8 seconds
- Narrative coherence: character state, position, and objects between shots must connect logically
- Narrative arc: setup → conflict → resolution/twist
- 8 elements: subject + scene + action + style + camera + composition + atmosphere + audio
Constraint priority (execute in this order when conflicts occur)
- Total word count ≤ 2000
- Narrative coherence + complete narrative arc
- Valid duration (15-second fast cuts 1-3 seconds; conventional narrative 4-8 seconds)
- Complete 8 elements
Three-Stage Workflow
| Stage | Input | Output |
|---|---|---|
| Idea | Story idea | Planning table → text prompt |
| Optimize | Shot + issue | Optimized prompt |
| Simplify | Shot prompt | Condensed version |
Stage 1: Idea (Seed)
Input: one-sentence story idea Output: storyboard planning table → complete 8-element prompt
1.1 Planning Stage
Based on the collected information, output the storyboard planning table:
| 镜头 | 类型 | 时长 | 主体 | 场景 | 动作 | 风格 | 镜头 | 构图 | 氛围 | 音频 |
|------|------|------|------|------|------|------|------|------|------|------|
| 1 | 建立 | 2秒 | 年轻女子... | 森林草地... | 躺卧苏醒 | 电影暖调 | 远景固定 | 三分法 | 晨光 | 鸟鸣 |
Table description:
- Type: setup / twist / resolution / climax
- Duration: allocated according to total duration (see table below)
- Subject: who, where, what appearance
- Scene: location, environmental props
- Action: specific behavior
- Style: visual aesthetics
- Camera: shot size + movement + angle
- Composition: rule of thirds/leading lines/center composition, etc.
- Atmosphere: lighting + mood
- Audio: music + sound effects + dialogue
Shot Allocation
Default strategy:Prioritize 15 seconds, then prioritize 7 shots (6-8 can be fine-tuned)。
| Priority | Shot count | Total duration | Allocation |
|---|---|---|---|
| 1 | 7 (default) | 15 seconds | 2+2+2+2+2+2+3 |
| 2 | 6 | 15 seconds | 2+2+3+3+3+2 |
| 3 | 8 | 15 seconds | 2+2+2+2+2+2+2+1 |
| 4 | 5 | 20-30 seconds | 4+4+5+5+4 |
| 5 | 4 | 18-24 seconds | 4+5+5+4 |
| 6 | 3 | 15-21 seconds | 5+5+5 (or 6+5+6) |
When unspecified, plan directly at 15 seconds / 7 shots; only change to 6 or 8 shots if the user explicitly requests it.
Chronological structure ("first... then... finally")
Use chronological order to make the action within a shot more coherent:
Emotional progression：
角色的表情从[不确定]逐渐转变为[越来越自信]，最后以[满足的成就笑容]结束。
Action sequence：
她先在[门口犹豫]，然后[深深吸一口气]，最后[坚决地推开门]。
Camera movement sequence：
场景以[广角定场镜头]开始，然后在第[3秒]平滑过渡到[中景]，最后以[特写]结束。
Time marker techniques：
- "At the [number]th second" is used to precisely control the timing of actions
- "Gradually transition from... to..." is used for smooth changes
- "End with..." is used for the closing action
The chronological structure is reflected in the "Action" column of the planning table to ensure that the action within each shot has a clear timeline.
Example: see examples.md
Stage 2: Refine
Input: shot + problem description Output: refined prompt
Problem → Refinement formula
| Problem type | Refinement keyword |
|---|---|
| Unnatural motion | + "Realistic physics controls all actions" + "Natural motion patterns" |
| Audio problem | + "Audio: [ambient sound]" + "No background music" |
| Ambiguous camera positioning | + "Camera is at [height] (thats where the camera is)" |
| Character inconsistency | + Detailed physical features |
| Quality problem | + "High resolution, crystal clear" |
| Poor lip sync | + "Clear lip sync" + colon format |
Output format
- Original problem (1 line)
- Refined prompt (8 elements)
- Change notes (list only additions/replacements)
Stage 3: Simplify
Input: over-refined prompt Output: simplified version
Simplification rules
- Remove redundant quality words: keep 1-2
- Delete repeated descriptions
- By default keep the core 8 elements (can be shortened into phrases)
- Keep camera positioning "(thats where the camera is)"
Simplification modes
- Standard simplification (default): keep the 8 elements, compress each element into 1 sentence
- Minimal feeding (only when the user explicitly requests it): compress to 2-4 elements (usually style/shot/action/audio)
8-element prompt structure
主体: [年龄、性别、服装、外貌、姿势、表情]
场景: [位置、建筑细节、道具、时间、天气]
动作: [具体动作、微表情、手势]
风格: [视觉美学、调色板]
镜头: [景别 + 运动 + 角度 + (thats where the camera is)]
构图: [三分法/引导线/景深]
氛围: [灯光：三点照明/黄金时刻/霓虹等]
音频: [对话(冒号格式)、环境音、音乐]
Key techniques
- Camera positioning：
相机位于[高度](thats where the camera is)or a similar positioning description - Dialogue format: use colon format to prevent subtitles
女人说：'台词' - Subject details: for regular shots, 8-15 physical features are recommended; for short shots, keep 4-8 key features
Detailed techniques:
- camera.md - Camera movement, angle, shot size
- composition.md - Composition techniques
- lighting.md - Lighting atmosphere
- style.md - Color palette
- audio.md - Dialogue, ambient sound, music, sound effects
Usage method
/screenwriter "女孩在森林醒来，变成蝴蝶飞走"
/screenwriter 优化 [镜头提示词] + [问题描述]
/screenwriter 简化 [镜头提示词]
Notes
- Total volume control: Storyboard ≤ 2000 characters
- Complete 8 elements = master level
- Adjust camera positioning syntax according to the platform
- Negative prompts: it is recommended to add "no watermark, no subtitles, no text overlay"
- Self-check before each output: whether duration, word count, 8 elements, and narrative coherence are all satisfied