Text Mode vs Sketch Mode: When to Use Each

CloudDiagram.ai gives you two distinct paths to a professional architecture diagram. Both produce the same interactive, editable output — but they start from very different inputs. Here is how to choose.

Text Mode: Start from an Idea

Text mode is ideal when you know what you want but have not drawn anything yet. You type a description like:

"Serverless event processing pipeline with API Gateway, SQS queue, Lambda consumers, DynamoDB for storage, and CloudWatch for monitoring."

The AI asks 3–5 follow-up questions to clarify your intent — things like which database engine you prefer, whether you need multi-AZ, or how many Lambda functions are involved. Then it generates the full diagram.

Best for:

• Greenfield designs where you are exploring options
• Quickly iterating on architecture ideas during planning meetings
• Generating standard patterns (three-tier, serverless, microservices)
• People who think in words rather than pictures

Sketch Mode: Start from a Drawing

Sketch mode shines when you already have something visual — a whiteboard photo, a napkin sketch, or even a screenshot from another tool. Upload the image and the AI identifies every component, label, and connection, then converts it into a clean diagram with real AWS icons.

Best for:

• Preserving whiteboard session outputs as proper documentation
• Converting hand-drawn mockups into presentation-ready diagrams
• Digitizing existing paper-based architecture drawings
• Reverse-engineering a diagram from a screenshot

How the AI Processes Each Mode

Text Mode Pipeline

1. You submit a description
2. AI generates clarifying questions
3. You answer questions
4. AI generates a complete diagram JSON with nodes, edges, and groups
5. The canvas renders the interactive diagram

Sketch Mode Pipeline

1. You upload an image (with optional context)
2. AI uses vision capabilities to analyze every element in the image
3. AI generates clarifying questions specific to what it detected
4. You answer questions
5. AI generates the diagram JSON matching your sketch
6. The canvas renders the interactive diagram

The sketch pipeline costs 3 credits vs 1 for text because it involves an additional vision analysis step that requires more compute.

Can I Mix Both?

Currently, each generation uses one mode. However, a powerful workflow is:

1. Start with text mode to generate an initial diagram
2. Edit the diagram interactively on the canvas
3. Export as .drawio for further modifications
4. Take a screenshot of a modified version
5. Use sketch mode to regenerate with new changes

This hybrid approach lets you leverage both AI speed and human judgment.

Tips for Better Results

Text mode tips:

• Be specific about AWS services (say "Aurora PostgreSQL" not just "database")
• Mention networking requirements (VPC, subnets, AZs)
• Include security components (WAF, IAM, security groups)
• State the pattern name if applicable ("three-tier", "microservices", "event-driven")

Sketch mode tips:

• Use clear, dark lines on a light background
• Label your components legibly
• Draw arrows to show data flow direction
• Group related services visually (even rough circles help)
• Add context in the text field to tell the AI what your sketch represents