Adding Nodes

Nodes are the building blocks of workflows. Each node performs a specific operation — loading a model, generating noise, applying conditioning, denoising latents, and more. By adding nodes to the canvas and connecting them together, you create a complete image generation pipeline.

Opening the Node Picker

The node picker is a searchable command palette that lists every available node. There are three ways to open it:

Keyboard Shortcut

Press Shift + A or Space while the workflow editor is focused.

Add Node Button

Click the + button in the top-left corner of the canvas.

Drag from a Port

Drag a connection from any input or output port and release it over empty canvas. The picker will open with results filtered to compatible nodes only.

Finding a Node

When the node picker opens, you can immediately start typing to search. The search is fuzzy and matches against several properties of each node:

• Title — the display name (e.g. “Denoise Latents”)
• Type — the internal identifier
• Description — a short summary of what the node does
• Tags — category keywords
• Node Pack — the origin module (e.g. invokeai for built-in nodes, or a community pack name)

Each entry in the picker shows:

• A classification badge indicating stability — Stable, Beta (yellow), Prototype (red), or Special (green)
• The node title and node pack name
• A brief description

Click a node or press Enter to add it to the canvas. The node will be placed near the center of your current viewport, or at your cursor position if you opened the picker by dragging from a port.

Tip

If you opened the picker by dragging from a port, the list is automatically filtered to show only nodes that have a compatible input or output for that connection. This is a fast way to discover which nodes work together.

Special Nodes

In addition to invocation nodes (which perform image generation operations), the picker includes two special utility nodes:

Notes

A sticky-note text area for documenting your workflow. Useful for leaving yourself reminders or explaining sections of a complex graph to others.

Current Image

Displays the current image being generated or the most recent output. Helpful for monitoring progress in long workflows.

Connecting Nodes

Nodes have input ports on their left edge and output ports on their right edge. Ports are color-coded by data type so you can quickly identify compatible connections.

1. Drag from an output port on one node toward the canvas.
2. Drop onto a compatible input port on another node. Compatible ports will remain highlighted; incompatible ports will appear greyed out.
3. A bezier edge is drawn between the two ports, representing the data flow.

Note

You can also drop a connection onto a node without targeting a specific port. InvokeAI will automatically connect to the first compatible port it finds on that node.

Connection Rules

• Connections must be between compatible data types (matching colors).
• A node cannot connect to itself.
• Each input port accepts only one connection (but an output can connect to many inputs).
• Connections snap within a 30px radius of a port for easy targeting.

Reconnecting and Removing Edges

• Reconnect an edge by dragging it from its current port to a new one.
• Remove an edge by dragging it away from its port and releasing it over empty canvas.
• Delete selected edges with Delete or Backspace.

Connectors

Connectors are small editor-only nodes that exist purely to reroute edges for a cleaner-looking graph. They are saved with the workflow but are flattened out of the graph before execution, so the runtime never sees them — you cannot use them to add logic, only to tidy wiring.

Ways to add a connector:

• Right-click empty canvas → Add Connector, then drag connections to and from it.
• Double-click an existing edge to insert a connector at that point, splicing it in.

Other behaviors worth knowing:

• Target-first wiring works. You can connect a connector’s output to a downstream target field before hooking up its upstream source. The connector stays unresolved until a compatible source is connected; incompatible upstreams are rejected.
• Type compatibility is enforced through the connector, exactly as for normal edges.
• Deleting a connector splices through any edges that pass through it:
  • 1 → 1: the source is reconnected directly to the target.
  • 1 → N: the source is reconnected to every compatible downstream target.
  • 1 → 0: the connector is removed, no edges created.
  • If a splice-through would produce an invalid graph, Delete Connector is disabled.
• Connectors persist across workflow save / load.

Configuring Nodes

Once a node is on the canvas, you can configure it by editing its input fields directly. Each node exposes a set of fields specific to its function — for example, a noise node has a Seed field, while a model loader has a Model selector.

• Inline editing — Click on any input field to edit its value directly on the node.
• Renaming — Right-click a node’s title or any input label to rename it.
• Use Cache — Toggle the caching option in the node footer to reuse previously computed values and speed up repeat runs.
• Collapse — Click the collapse button on the node header to minimize it, keeping the canvas tidy.

Managing Nodes

Use these shortcuts to work efficiently with nodes on the canvas:

Action	Shortcut
Add Node	Shift + A or Space
Copy	Ctrl/Cmd + C
Paste	Ctrl/Cmd + V
Paste with Edges	Ctrl/Cmd + Shift + V
Select All	Ctrl/Cmd + A
Delete	Delete or Backspace
Undo	Ctrl/Cmd + Z
Redo	Ctrl/Cmd + Shift + Z
Select Multiple	Shift + Click & Drag

Tip

All keyboard shortcuts are customizable. Open the Hotkeys modal with Shift + ? to view or change any binding.

Adding to Linear View

Any input field on a node can be promoted to the Linear View, which provides a simplified UI for your workflow — perfect for sharing with others or for quick iteration.

1. Right-click on an input label on any node.
2. Select “Add to Linear View”.
3. The input now appears in the Linear View panel, where you can adjust it without navigating the full graph.

Custom names you set on input fields will carry over into the Linear View.

Installing Community Nodes

InvokeAI’s node system is extensible. Community-created nodes can add new capabilities to your workflows — from specialized image processing to LLM-powered prompt generation.

The easiest way to install a community node pack is through the Custom Node Manager: paste a Git URL in the Nodes sidebar tab and the pack is cloned, loaded, and made available without a restart.

If you prefer to install manually:

1. Find a node pack from the Community Nodes list.
2. Clone or download the node pack into the nodes folder inside your InvokeAI installation directory.
3. In the Custom Node Manager, click Reload (or restart InvokeAI). The new nodes will appear in the node picker.

Note

git clone is preferred over downloading a ZIP — it makes it easy to update node packs later with git pull.

For more details and a full catalog of available community nodes, see the Community Nodes page.