Flows is a visual, node-based canvas for building multi-step AI generation pipelines. Instead of generating an image, then manually carrying it to another tool to turn it into a video, then another tool to add voiceover, Flows gives you a single canvas to build, run, and iterate on your entire AI generation pipeline. Connect a Text Prompt to an Image Generator, to a Video Generator, and run the whole thing, or just from any point forward, without losing what came before.
Use it to test multiple AI models side by side, compare whether a text prompt or a reference image gets you closer to your vision, or build a repeatable workflow you can reuse and refine over time.
Key Features
- Node-based canvas: An infinite pan-and-zoom workspace where each node represents a generation step or an input.
- Typed connections: Nodes connect through color-coded ports (text, image, video, audio) — you can only connect compatible ports, so it's clear what feeds into what.
- Run a whole flow: Pressing Run executes the entire connected chain of nodes in the right order, showing live status on every node as it goes.
- Multi-output galleries: Every result a node produces is kept, not overwritten, so you can compare outputs and pick the one that flows downstream.
- Auto-saving canvases: Your canvas — nodes, connections, positions, and settings — saves automatically and restores exactly as you left it.
- Example Project: A ready-made canvas with sample flows already generated, so you can see how Flows works before building your own.
Who it's for
Flows is built for creators and marketers who regularly combine several AI generation steps into one piece of content — for example, going from a text prompt to a finished image to an animated video — and want a single, reusable, visual record of how that content was built.
How to access Flows
- Click on Flows in the Navigation bar of the homepage.
How to get started from scratch
- Select + New Flow from the Flows home page
- Select your first node from sidebar
- Text prompt
- Prompt assistance (write a simple prompt that will be improved for you)
- File uploader (to use as image reference)
- Image generator
- Video generator
- Add the relevant instructions or settings into your Node (dependent on the model selected)
- Drag from a node's output port to a compatible input port on another node. Ports are color-coded by type (text, image, video, audio), so it's easy to see what can connect to what.
- If you drag a connection and drop it on empty canvas instead of onto another node, a searchable node picker opens — showing only node types that can accept that connection. Choosing one creates and connects that node in a single step.
- Each node's in put settings are shown inline and change based on the model you select for that node.
- Deleting a node also removes any connections to and from it.
- To change the connection hover over the chain click the scissors
4. Keep adding nodes until you finish the flow, altering the inputs of each node as a new node is created.
5. When you are ready select Run on the bottom right of each node and select from:
- Run node - Use credits to generate that specific node
- Run from here - Use credits to generate from this node forward
- Run full flow - Use credits to gerbate the whole flow
Note: Before running, each generator node shows its cost in credits, so you know what a run will cost before you start it.
Note: If a node fails, you'll see a specific error message directly on that node, along with a Retry button to re-run just that node — you don't need to reconfigure or re-run anything upstream.
How to get started with a Template
- Select Templates from the Flows home page
- Browse the available templates and click on one to use it
- A copy of the template opens automatically in your canvas
- Read the sticky notes on the canvas for guidance on what inputs each node needs
- Update the relevant nodes with your own content (e.g. upload an image, edit the prompt)
- Select Run on the bottom right of a node and choose Run node, Run from here, or Run full flow
Node types
- Prompt (Inputs) — a freeform text field for your prompt.
- Prompt Assistant (Inputs) — takes text and/or an image and uses an LLM to expand or transform it into text.
- File Uploader (Inputs) — upload your own image, video, or audio, or pull one in from your library.
- Image reference
- Video inference
- Start frame
- End frame
- Image Generator (Generators) — generates an image from text and/or image input, with settings for model, aspect ratio, resolution, and number of images per run.
- Video Generator (Generators) — generates a video from up to three image inputs (start frame, end frame, reference), plus text and audio, with settings for model, aspect ratio, resolution, and duration.
Note: Asset inputs and the available amount, vary depending on the settings of the node.
Note: Click the 3 black dots on any node to:
- Duplicate
- Rename
- Download
- Delete
Toolbar
- Keep still
- Move
- Sticky notes - create a sticky note to add notes about your project
- Delete
- Zoom in and out
- Zoom to selected area
Working with results
- Every time you re-run a node, its new result is added to that node's results gallery — nothing is overwritten.
- Once a node has more than one result, its preview switches from a single large image to a grid of thumbnails.
- Click any thumbnail to make it the node's active result — that's the version downstream nodes will use when the flow runs.
- Hover over any result to favorite or download it.
💡 Tip: Start from a Template card if you're new to Flows — it comes pre-populated with sample prompts, inputs, and already-generated results, so you can see a working pipeline before building your own.
💡 Tip: Use "run from a node" to test one step at a time while you're building a flow, then run the full chain once everything's connected the way you want.
Commonly asked questions
Does my project save automatically?
Yes. Your canvas — including node positions, connections, and settings — saves automatically with no explicit save step, and restores exactly as you left it when you reopen it.
Are previous results kept when I re-run a node?
Yes. Every result a node produces is kept in that node's results gallery rather than being overwritten, so you can compare outputs and choose which one to use.
Can I see the cost of a run before I run it?
Yes. Each generator node shows its per-run cost in credits before you run it.
What happens if a node fails?
The failing node shows a specific error message directly on the node, along with a Retry action to re-run just that node.