# Nodes & connections Each node has a job. A connection passes one step's output into the next. Source: https://www.creativly.ai/docs/flow/nodes-and-connections ![Dark Flow canvas at normal zoom: a Text brief connected to a completed Nano Banana 2 image, with node settings, instructions and Actions visible.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-brief-workspace-e2f2584aa20fddb1.webp) The Text node feeds the Image node. Select a step to see its settings and actions. ## Choose the right starting point Right-click an **empty area of the canvas** to open the node picker. Choose **Prompt** for text you write yourself, or a generation node for something a model should make. ![Flow's actual node picker, including LLM, Image, Video, Set, Prompt and media-import options.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-node-picker-2065cdf0ec9ffdfe.webp) Add a step where you need it. Imported media is a source; a generation node creates a new result. | Node | What belongs here | | -------------------------------- | --------------------------------------------------------------------------------------------------------------- | | **Prompt / Text** | Your brief, reusable instructions, or notes. It does not call a model. | | **LLM** | Instructions for AI-written text, such as refining a brief or describing connected media. | | **Image** | A new image or an edit guided by connected references. | | **Video** | A generated clip or a supported video edit. | | **Audio** | [Spoken lines or narration](https://www.creativly.ai/docs/flow/text-to-audio), using a script and chosen voice. | | **Set** | Several variants, shots, or sequences with shared direction. | | **Import Image / Video / Audio** | An existing file to use as an input. | Model and provider choices determine which operations and inputs are available. ## Connect an output to an input Drag from the small **output port on the right** of a source node to a compatible **input port on the left** of the next node. In the example, **Text → Image** passes the campaign brief into image generation. The image node's own instructions add the specific scene. Both reach the model. Moving cards near one another does not connect them. Leave some space between steps so their curves and ports remain easy to follow. ## Match the input to the job Hover an input port to read its name. The available ports change with the selected model; you do not need to fill every one. ![Actual video node with the Reference image tooltip open, separate image and video connections, Gemini Omni 1.1 Flash settings, and Media 2 below the edit prompt.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-input-role-port-215cde2c245a6fb0.webp) Image reference (blue) + source video (amber) → a separate edit. | Input | Use it for | | ------------------- | -------------------------------------------------------------- | | **Prompt** | Text from a connected brief or LLM result. | | **First frame** | The starting image of a new clip. | | **Last frame** | Its ending image, together with a first frame. | | **Reference image** | Visual guidance, such as the subject or setting. | | **Reference video** | A clip to edit, or video guidance where the model supports it. | For the [edit shown here](https://www.creativly.ai/docs/flow/edit-video), use **Reference video + Reference image**. For [image → film](https://www.creativly.ai/docs/flow/image-to-video), use **First frame**. Keep a first/last-frame setup separate from image/video references for the same run. The main Prompt Bar has [matching reference chips](https://www.creativly.ai/docs/flow/prompt-bar#choose-video-inputs), so you can also choose these roles without dragging a connection. ## Select a node to work on it Click a node to reveal its **model and output settings above**, and its **instructions and Run control below**. The right-hand **Actions** panel changes with the selected media. ![Actual image quick actions: Edit Image, To Video, Describe Image, Batch, Import Image, Remove Background, Erase and Expand.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-image-actions-9fc5b470e58b3814.webp) Actions for an image. To Video adds the next connected step; Edit Image starts a new image edit. The **model** chooses what makes the result. **Run nodes with** sets the default provider for nodes; an individual node can override it. [Check the model and provider](https://www.creativly.ai/docs/models-and-credits#choose-how-to-run) before starting a job. [Try a real image → cutout example](https://www.creativly.ai/docs/flow/remove-background): add the connected step, select Remove Background, and keep the transparent PNG. ## Decide how much to run The arrow beside **Run** shows the scope and estimate: * **Run from here:** this node and its downstream runnable steps. * **Run till here:** this node and the connected steps leading into it. ![Run scope menu: Run from here is one credit and one node; Run till here is one credit and two nodes because the upstream Text node is not a generation.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-brief-run-scope-517e132c68af20a6.webp) This graph has two nodes, but only the image costs credits. The Text node is a reusable input. Changing a prompt does not update an old output by itself. Run the step you want to regenerate, and check its result before continuing downstream. [Run a selection and find earlier results →](https://www.creativly.ai/docs/flow/groups-and-runs) [Brief → product image](https://www.creativly.ai/docs/flow/brief-to-image)Build the two-node example shown here. [Image → product film](https://www.creativly.ai/docs/flow/image-to-video)Give a completed still a little motion. Open in Creativly: [Flow ↗](https://www.creativly.ai/app/flow)