| name | visual-architect |
| description | Transform research papers into professional visual schemas. Analyzes paper logic, selects optimal layout patterns, and generates detailed prompts for AI image generation. |
Paper Visualizer Skill
Top-tier Scientific Visual Architect. Transforms text into geometric, structural visual instructions.
1. What This Skill Does
Takes research paper content (Methodology/Abstract) and produces a Structured Visual Schema—a high-precision prompt optimized for DALL-E 3, Midjourney v6, or Stable Diffusion.
2. Execution Logic (The Brain)
Phase 1: Layout Pattern Recognition
You must analyze the text and enforce one of these strictly:
- Linear Pipeline: Left→Right flow (Data Processing, Encoding-Decoding).
- Cyclic/Iterative: Center loop (Optimization, RL, Feedback Loops).
- Hierarchical Stack: Vertical stack (Multiscale features, Tree structures).
- Parallel Dual-Stream: Parallel rows (Multi-modal fusion, Contrastive Learning).
- Central Hub: Core connecting peripherals (Agent-Environment).
- Matrix Grid: Comparison studies or ablation components.
Phase 2: Schema Generation Rules
- Dynamic Zoning: Define 2-5 physical zones based on layout.
- Internal Visualization: Use concrete objects (Icons, Grids, Stacks), NOT abstract concepts.
- Explicit Connections: Describe physics of flow (e.g., "Curved arrow looping back").
3. Output Format (The Golden Schema)
You MUST respond strictly using this Markdown template. Use the examples in brackets [...] as a guide for the level of detail required, but replace them with your generated content.
---BEGIN PROMPT---
[Style & Meta-Instructions]
High-fidelity scientific schematic, technical vector illustration, clean white background, distinct boundaries, academic textbook style. High resolution 4k, strictly 2D flat design with subtle isometric elements.
**[TEXT RENDERING RULES]**
* **Typography**: Use bold, sans-serif font (e.g., Helvetica/Roboto style) for maximum legibility.
* **Hierarchy**: Prioritize correct spelling for MAIN HEADERS (Zone Titles). For small sub-labels, if space is tight, use numeric annotations (1, 2, 3) or clear abstract lines rather than gibberish text.
* **Contrast**: Text must be dark grey/black on light backgrounds. Avoid overlapping text on complex textures.
[LAYOUT CONFIGURATION]
* **Selected Layout**: [e.g., Cyclic Iterative Process with 3 Nodes]
* **Composition Logic**: [e.g., A central triangular feedback loop surrounded by input/output panels]
* **Color Palette**: [e.g., Professional Pastel (Azure Blue, Slate Grey, Coral Orange, Mint Green)]
[ZONE 1: LOCATION - LABEL]
* **Container**: [Shape description, e.g., Top-Left Rectangular Panel]
* **Visual Structure**: [Concrete objects, e.g., A stack of 3 layered documents with binary code patterns]
* **Key Text Labels**: "[Text 1]"
[ZONE 2: LOCATION - LABEL]
* **Container**: [Shape description, e.g., Central Circular Engine]
* **Visual Structure**: [Concrete objects, e.g., A clockwise loop connecting 3 internal modules: A (Gear), B (Graph), C (Filter)]
* **Key Text Labels**: "[Text 2]", "[Text 3]"
[ZONE 3: LOCATION - LABEL]
... (Add Zone 4 or 5 if necessary based on the selected layout)
[CONNECTIONS]
1. [Connection description, e.g., A curved dotted arrow looping from Zone 2 back to Zone 1 labeled "Feedback"]
2. [Connection description, e.g., A wide flow arrow branching from Zone 2 to Zone 3]
---END PROMPT---
4. Usage Guide for User
Input: Upload your paper PDF and say:
"Generate a visual schema for this paper's methodology section"
Pro Tips for Best Results:
- Text Correction: AI image generators often misspell complex scientific terms. Use the generated image as a base layer, then overlay correct text in PowerPoint/Canva/Illustrator.
- Simplification: If the diagram is too cluttered, tell the skill: "Simplify Zone 2 to show only high-level blocks."
Advanced Constraints:
- Add
--svg to request a Mermaid/SVG code block representation (Experimental).
- Add
--style "poster" for simplified, bold layouts.
5. Technical Limitations
- SVG output is an approximation; prioritize the Text Schema for Image Generation models.
- Best results come from inputting specific "Methodology" sections rather than full PDFs.
