Workflows is Layer's feature for chaining together multiple AI steps into a single automated flow. Instead of generating one asset at a time, you can set up a recipe once and run it as many times as you need, on demand.

A workflow is a chain of nodes that transform data in a specific order. Think of it like a recipe: you gather your raw ingredients (prompts and reference images), prepare them with generator nodes into images, videos, or audio, then "cook" those assets further using nodes like Edit Image, the Composition Node, or Upscale until you reach the final result.

Some workflows are simple, like turning a single prompt into a single image. Others are complex, generating multiple images, arranging them into a scene at a specific resolution, then animating the result into a video.

Workflows also scale. A single workflow can take in many different inputs and produce many outputs at once, a "many to many" relationship. This lets you do things like change the pose of 50 characters in one run, or generate 15 video variants for a campaign in a single pass.

Access Workflows from the left sidebar of your workspace or from the Workflows tile on the home page. We recommend starting with one of Layer's Featured Workflows to see design patterns in action before building your own.

To use a featured workflow, import it into your workspace first. This creates a local, editable copy. From there, click Prepare Workflow to review or update any prompts and reference images, then run it. Workflows can take anywhere from a few minutes to, for especially complex or large runs, up to an hour to complete, since partner model providers need time to render and deliver every asset the workflow calls for.

Once complete, your generated assets appear grouped by completion time.

Filtering and finding workflows

Use the sidebar on the Workflows page to quickly find the workflow you need.

Search workflows
Type a keyword into the search bar at the top to find workflows by name.

View filters

Label
Filter workflows by tags that you create. Use the search icon to find a specific label, or the + icon to create a new one for organizing your own workflows.

Created By
Filter workflows by who created them. By default this shows You, with the option to show more to reveal other workspace members and view workflows they've created.

Status
Filter by publishing state:

Visibility
Filter by who can see the workflow:

Selecting one or more workflows

You can select more than one workflow at a time to apply actions in bulk. Click a workflow thumbnail while holding Cmd (Mac) or Ctrl (Windows) to add it to your selection. Selected workflows are highlighted with a purple border and checkmark.

Once you've selected one or more workflows, a toolbar appears showing how many are selected along with the following actions:

Right-click menu actions

Right-click (or click the share icon) on any workflow thumbnail to open a quick-actions menu with the following options:

Most people using Workflows do not need the full editor open. Once a workflow is in a good state, use the Run Workflow button, which only asks you to fill in the specific inputs you want to change and skips the graph view entirely. Every run keeps the normal Layer conveniences: you can rename the session, browse past Workflow sessions, and the form remembers changes you made during the session.

To make changes to an existing workflow, duplicate it from the three dot Settings menu, or open it directly and start a new draft. Starting a draft keeps your changes local so they don't affect the published version other people may be using.

Inside a draft, you can test your changes two ways:

When you're happy with your draft, publish it back to your workspace, or save it as a new workflow if others are relying on the existing published version. Publishing over an existing workflow updates it for everyone using it.

Workflow Inputs are the nodes that feed prompts, reference images, and reference videos into the rest of your graph. Every input can be either static or dynamic:

You can add as many inputs as you like. This is one of Workflows' most powerful features: once you have a workflow set up to change a character's pose, for example, you can feed in as many characters as you want and get the same treatment applied to all of them.

Adding inputs multiplies your generations. If a workflow has one 2D generator fed by 3 prompts, that generator runs 3 times; if it produces 4 images each time, you get 12 images total from those 3 prompts.

Workflow inputs aren't limited to generators either. A Reference Image can feed a 2D Generator or a Composition Node. A Reference Video can feed a Stitch Video node or a video model that accepts a base video. Every input needs a default value (text, image, or video) before the workflow can run; an empty input node puts the whole workflow in an invalid state. When someone runs your workflow, they'll see each input in its default state (the state you last left it in) and can replace or add to it from there.

Complex workflows often combine static and dynamic inputs to keep the number of things a user has to review before running the workflow to a minimum.

For details on the individual nodes you can use inside a workflow, see: