π¨ AI Service that Generates Professional Diagrams and Visual Designs from Natural Language
Based on LLM 'Gemma-3-R1984-27B' Model & FLUX Image Generation Model. Powered by VIDraftβ‘
Ranked 2nd on Google DeepMind FACTS Grounding Leaderboard
Service Link: https://huggingface.co/spaces/openfree/Chart-GPT
π Overview
ChartGPT is an AI-based service that automatically generates professional diagrams and visual designs from Korean and English prompts. Without needing to learn complex diagramming tools, simply describe what you want and AI will automatically generate structured diagrams or beautiful visual designs.
π Core Technology
Base LLM: Gemma-3-R1984-27B
Achieved 2nd place on Google DeepMind FACTS Grounding Leaderboard(https://huggingface.co/spaces/MaziyarPanahi/FACTS-Leaderboard)
State-of-the-art language model with exceptional factuality and accuracy
Outstanding performance in both Korean and English
Optimized for complex structuring tasks
Image Generation: FLUX.1-schnell
High-quality image generation based on latest diffusion models
Fast generation speed with high quality
Support for various styles and concepts
MCP (Model Context Protocol) Support
Standardized AI Integration: Full support for Anthropic's MCP protocol
Integration with Various AI Tools: Seamless integration with major AI services like Claude and ChatGPT
Extensible Architecture: Operates as an MCP server for easy connection with other applications
β¨ Key Features
1. Support for 5 Diagram Types
πΊοΈ Concept Map: Visualize relationships between topics and related concepts
π Synoptic Chart: Display the overall structure of processes or systems at a glance
βοΈ Radial Diagram: Express structures expanding radially from a central concept
π Process Flow: Represent workflows or decision-making processes sequentially
π WBS Diagram: Display projects broken down hierarchically
2. Support for 6 Visual Design Types (NEW)
π¨ Product Design: Industrial design concepts and product sketches
π§ Mindmap: Visually rich thought maps
π± Mockup: UI/UX wireframes and app designs
π Infographic: Data visualization and information design
π Diagram: Business workflows and process visualization
π Flowchart: Decision flows and process charts
3. AI-Based Automatic Generation
Generate diagrams from natural language prompts
Automatic translation from Korean to English (for Design Types)
Automatic structured JSON generation
Optimized prompt templates for each design type
4. Brave Search Integration
Reflect latest information through real-time web search
Generate more accurate and richer content
Optional activation
5. Flexible Output Formats
PNG format (high-quality images)
SVG format (vector graphics, infinitely scalable)
π MCP (Model Context Protocol) Utilization
ChartGPT supports MCP for integration with various AI applications:
MCP Use Cases
AI Agent Integration: AIs like Claude and GPT can directly request diagram generation
Workflow Automation: Automatically insert diagrams during document creation
Multimodal AI Pipeline: Automate text analysis β diagram generation
π How to Use
Using the AI Assistant Tab
1. Enter Prompt
Describe the diagram or design you want in natural language
Diagram example: "Create a flowchart showing the machine learning process"
Design example: "smartphone banking app design"
2. Select Type
Diagram Type: Generate structured diagrams (JSON-based)
Design Type: Generate visual designs (FLUX-based, English only)
3. Use Brave Search (Optional)
Activate checkbox when latest information is needed
Web search results will be reflected in generation
4. Click Generate Button
AI automatically generates structure
Diagram or design image automatically rendered
Direct JSON Input
Fine-tune by directly editing JSON in each diagram type tab
π― Use Cases
Business
Project planning (WBS)
Business process documentation (Process Flow)
Organizational chart creation (Radial Diagram)
Product concept design (Product Design)
Marketing infographics (Infographic)
Education
Learning content organization (Concept Map, Mindmap)
Course explanatory materials (Synoptic Chart)
Research topic structuring (Concept Map)
Educational material visualization (Diagram)
Technical Documentation
System architecture diagramming
API flow explanation
Data processing pipeline visualization
UI/UX mockups (Mockup)
Technical flowcharts (Flowchart)
π§ Advanced Configuration
Customizing JSON Structure
Each diagram type has its unique JSON structure:
Concept Map Example:
json{
"central_node": "Artificial Intelligence",
"nodes": [
{
"id": "ml",
"label": "Machine Learning",
"relationship": "includes",
"subnodes": []
}
]
}
Prompt Writing Tips
Be Specific: "AI classification system" β "AI classification system including deep learning, machine learning, and natural language processing"
Specify Relationships: "A is part of B", "X leads to Y", etc.
Describe Hierarchical Structure: "A at the top level, with B and C below", etc.
For Design Types: Input in Korean is automatically translated to English
π‘ Model Performance
Gemma-3-R1984-27B Features
High Factuality
Structuring Ability: Accurately understands complex relationships and hierarchical structures
Multilingual Support: Ensures equal quality in both Korean and English
Context Understanding: Accurately grasps prompt intent
FLUX.1-schnell Features
High-Quality Images: Professional-level visual design generation
Fast Generation: Efficient inference for quick results
Diverse Styles: Support for sketches, wireframes, infographics, and various styles
π οΈ Tech Stack
Frontend: Gradio (with built-in MCP server)
AI Models:
Gemma-3-R1984-27B
FLUX.1-schnell (High-quality image generation)
Diagram Engine: Graphviz
Search Engine: Brave Search API
Protocol: MCP (Model Context Protocol)
Language: Python