/alibabacloud-video-prompt-architect

name: alibabacloud-video-prompt-architect
description: |
  Generates structured, high-quality prompts for AI video and image generation models.
  Transforms natural language descriptions into optimized prompts adapted for 18 models including Happy Horse, Seedance, Kling, Pika, Midjourney, Recraft, FLUX, and more.
  Use when creating video prompts, image prompts, product images, posters, or adapting prompts across different AI generation models.
  Triggers: "生成视频提示词", "视频prompt", "文生视频", "图生视频", "文生图", "商品图", "海报生成", "AI生成提示词", "prompt architect", "media prompt"

AI Media Generation Prompt Architect

Automatically decompose a user's natural language request and generate high-quality structured prompts adapted for different AI models.

Applicable Scenarios

• Users without prompt experience can quickly generate professional-level prompts
• Enterprise operations, design, and short-video teams for batch content production
• Unified prompt generation entry point for AI creation platforms
• Cross-platform multi-model creative adaptation

Architecture

User natural language input → Intent analysis → Structured decomposition → Mode adaptation → Self-review reflection → Output high-quality Prompt

Components: `Intent Parsing Engine` + `Structured Prompt Generator` + `Multi-Mode Adapter` + `Negative Prompt Generator` + `Quality Self-Checker`

---

Core Workflow

1. Receive User Input

Accept the user's natural language description, which can be a brief sentence, e.g.:

• "A girl dancing under cherry blossom trees"
• "A high-tech smartphone product showcase"
• "A romantic wedding photo at sunset by the sea"

2. Intent Analysis and Generation Mode Determination

Based on user input, automatically determine the most suitable generation mode and select the target model according to user specification or default strategy:

Task Type Identification

| Trigger Keywords | Generation Mode | Default Model | |------------------|-----------------|---------------| | Video, animation, motion, dynamic, clip | Text-to-Video | Happy Horse | | Reference image, multi-character, image fusion video | Reference-to-Video | Happy Horse r2v | | First frame, image-to-video, animate image | Image-to-Video | Happy Horse i2v | | Image, photo, illustration, wallpaper, concept art | Text-to-Image | Nano Banana | | Product, merchandise, e-commerce, showcase | Product Image Generation | seedream | | Poster, promotion, advertisement, banner | Poster Generation | Midjourney |

Supported Video Generation Models

| Model | Use Case | Language | Prompt Style | |-------|----------|----------|--------------| | **Happy Horse** | Business custom videos, platform integration, lightweight creativity | **Chinese (mandatory)** | Structured Chinese, camera terms directly usable | | **Seedance** | Short videos, narrative segments, motion shots | **Chinese (mandatory)** | Structured Chinese: subject + action + scene + camera + style | | **Kling** | Cinematic videos, ads, drama segments | **Chinese (mandatory)** | Structured Chinese, emphasizing camera movement and visual texture | | **Wanx** | Chinese-native video, general content, marketing videos | **Chinese (mandatory)** | Chinese structure, reduce abstract words | | **Veo** | High-quality video, commercials, cinematic visuals | English | Complete description, emphasizing atmosphere, rhythm, scene details | | **Sora** | Complex narratives, multi-character, physical consistency | English | Semantically coherent, clear subject relationships and action logic | | **Hailuo** | Short videos, social content, quick production | **Chinese (mandatory)** | Concise and direct, emphasizing action, emotion, and style | | **Runway** | Creative ads, brand content, stylized videos | English | Natural language + style directives, emphasizing brand consistency and visual style | | **Pika** | Social media shorts, creative effects videos, viral content | English | Concise and dynamic, emphasizing creative transitions and special effects (melt/inflate/explode/crumble) |

Supported Image Generation Models

| Model | Use Case | Language | Prompt Style | |-------|----------|----------|--------------| | **Nano Banana** | Quick image generation, creative exploration, lightweight visuals | English | Concise and efficient: subject + style + composition | | **GPT Image** | General images, design sketches, marketing graphics | English | Natural language, semantically clear | | **Grok Image** | Creative images, social media visuals, personalized graphics | English | Natural description, emphasizing effects and themes | | **seedream** | Commercial images, posters, high-quality pictures | **Chinese (mandatory)** | Structured Chinese: subject + scene + lighting + texture + composition | | **Qwen Image** | Chinese design needs, general images, marketing visuals | **Chinese (mandatory)** | Chinese-organized, emphasizing purpose and style | | **Midjourney** | Posters, concept art, stylized illustrations | English | Style + composition + material + lighting keyword combinations | | **FLUX** | API integration, developer workflows, self-hosted | English | Structured English: subject + scene + style + quality, precise control | | **Ideogram** | Posters/ads/covers, text rendering, typographic images | English | Natural language + text content directives, emphasizing layout and readability | | **Recraft** | Vector graphics, brand design, icons/logos, print materials | English | Design-oriented: subject + style + color palette + output format, supports SVG/EPS vector output |

> If the user does not specify a model, use the default model based on task type. Users can switch at any time. > For detailed model adaptation rules, see `references/generation-modes.md`

3. Structured Prompt Generation

Decompose user input into the following **8 structured components**, each generated independently:

3.1 Subject Description

• Clearly define the subject's appearance, features, and state

-