Building Workflows
Learn how to design and build effective AI workflows. From simple pipelines to complex multi-model systems.
Workflow Basics
A workflow is a connected graph of nodes that defines how data flows and transforms through your AI pipeline. Think of it as a visual program where:
- •Nodes are functions that perform specific tasks
- •Connections are data pipes between nodes
- •Execution follows the data flow from inputs to outputs
Common Workflow Patterns
Linear Pipeline
The simplest pattern: data flows in one direction from input to output.
Example: Text prompt → Image generation → Image preview
Multi-Step Pipeline
Chain multiple generations together, where each step builds on the previous.
Example: Prompt → Text generation → Image generation → Video generation
Fan-Out Pattern
One input feeds multiple parallel generations.
Example: One prompt → Multiple image variations with different models
Processing Pipeline
Apply tools and effects to generated content.
Example: Generate image → Crop to 16:9 → Adjust colors → Upscale 2x
Triggered Chain
Use trigger connections to control execution order.
Example: Generate text → Use text to generate image → Use image for video
Building Your First Workflow
Step 1: Define Your Goal
Before adding nodes, clarify what you want to achieve:
- •What content do you want to create?
- •What inputs do you need?
- •What transformations are required?
- •How will you use the output?
Step 2: Add Input Nodes
Start with input nodes to bring data into your workflow. Use Text Input for prompts, Image Input for reference images, or Video Input for source footage.
Step 3: Add Generative Nodes
Add the AI models that will create your content. Select appropriate models based on your quality, speed, and cost requirements.
Step 4: Connect Nodes
Click and drag from output ports to input ports. Connections show the data flow. Make sure port types match (text→text, image→image, etc.).
Step 5: Add Preview Nodes
Connect preview nodes to see your results. Preview nodes display generated content and let you download, expand, or copy outputs.
Step 6: Test and Iterate
Run your workflow, review results, and refine. Adjust prompts, change models, or add processing nodes until you get the desired output.
Workflow Organization
Keep your workflows clean and maintainable:
Use Descriptive Names
Rename nodes to describe their purpose (e.g., "Product Description" instead of "Text Generate").
Organize Spatially
Arrange nodes left-to-right (inputs → processing → outputs) or top-to-bottom. Keep related nodes grouped together.
Use Annotations
Add text annotations, sticky notes, or arrows to document complex workflows and explain decision points.
Minimize Crossings
Avoid connection lines crossing each other. Rearrange nodes to keep connections clear and easy to follow.
Lock Finalized Nodes
Lock nodes that are positioned correctly to prevent accidental movement while working on other parts.
Performance Tips
- •Use faster models for iteration: Start with quick models like GPT-4o Mini or FLUX Schnell, then upgrade to premium models for final outputs.
- •Test nodes individually: Run single nodes to verify they work before running the entire workflow.
- •Reuse successful outputs: Save generated content and use Image/Video Input nodes to reuse them in new workflows.
- •Monitor credit usage: Check the Info Panel to see estimated costs before running expensive workflows.