PhotoGPT Flow

PhotoGPT Flow is a visual canvas where you build AI generation pipelines by connecting nodes. Instead of switching between tools or copying outputs manually, you wire together the steps — image generation, video generation, text, media uploads — and run them as a single workflow.

Each node is a self-contained step: it takes inputs, runs an AI generation, and outputs results that feed directly into the next node. You can group nodes, save them as presets, and replay the entire pipeline whenever you need consistent output.

Open your PhotoGPT dashboard

What You Can Build With Flow

• Multi-step image pipelines — upload a reference image, generate a styled portrait, then enhance it further with additional nodes, all in one chain.
• Video scenes with control — pass first-frame and reference images into video generation to control exactly what the model builds on.
• Viral influencer content — create consistent avatars and scenes for social media by chaining custom models and reference images.
• Consistent brand campaigns — save a node group as a preset and reuse it across product launches or content series.

The Canvas

When you open Flow, you land on an empty canvas. This is your workspace where nodes live, connect, and run.

Navigating the Canvas

The Floating Dock

At the bottom of the screen is a dock with quick actions:

Node Types

Nodes are the building blocks of your workflow. Each one does one thing.

To add a node, press A or double-click anywhere on the canvas.

Selecting a Model

Each generation node (Image, Video, Text) needs a model to run. When you add a node, it defaults to a commonly used model, but you can change it at any time.

How to Select a Model

1. Click on the node to select it.
2. A settings panel opens on the right side of the canvas.
3. At the top of the panel, you'll see the current model name with a dropdown arrow.
4. Click the dropdown to see all available models grouped by provider.
   

Notes on Model Selection

• Different models produce different styles and quality levels — experiment to find what works best for your project.
• Not all models support all settings. If a model doesn't support a setting you've set, the panel will either hide that option or show a warning.
• Custom models are trained on your own images and reflect your personal style.

Adding and Connecting Nodes

Configuring Node Settings

Running a Single Node

You can run any individual node without running the entire workflow.

1. Click on the node to select it.
2. Look for the Run button at the bottom of the node.
3. Click Run to execute just that node.

The node will show a processing indicator while it runs, then display its output when complete. Running a single node is useful for testing or when you want to generate one step without running the full pipeline.

Saving Workflows as Presets

Once you have a group of nodes working well together, you can save it as a preset for reuse.

1. Select the nodes you want to save.
2. Click Save Preset in the actions panel.
3. Give the preset a name.
4. Later, press P on the dock and insert the preset — your node group loads instantly.

This is useful for workflows you run repeatedly: brand templates, product shot pipelines, or recurring video campaigns.

Connecting Reference Images and Text

You can connect external inputs to your generation nodes — images from your uploads or text content — to guide or fuel the generation.

Connecting Reference Images

When you connect an Asset node (containing an image) to an Image or Video node, the uploaded image influences the output:

• To an Image node — the image can be used as a style reference, composition guide, or input for editing.
• To a Video node — the image can be set as the first frame (starting shot), last frame (ending shot), or a reference that guides the overall style without locking the visual.

To assign a reference image role:

1. Connect an Asset node to your Image or Video node.
2. Click on the image pill that appears inside the target node.
3. Choose the role: First Frame, Last Frame, or Reference Image.

Only one image can be assigned to First Frame or Last Frame. You can have multiple Reference Images depending on the target node's model type.

Connecting Text

When you connect a Text node to an Image or Video node, the text content is passed as part of the prompt:

• The Text node outputs text content (headlines, descriptions, or copy).
• This text flows into the generation node's prompt, adding to what you've already typed.

To connect, drag from the Text node's output handle to the generation node's prompt input.

How Connections Flow

The flow of data is one-directional — from the output of one node to the input of another:

1. Source node generates or provides content (image, video, or text).
2. Connected downstream node receives that content as input.
3. The downstream node uses it alongside its own settings and prompt to produce output.
   

Undo, Redo, and History

• Undo — Press Ctrl+Z or use the undo button in the header
• Redo — Press Ctrl+Y or Ctrl+Shift+Z
• History is saved across sessions in your browser, so you can undo actions from a previous session.

Keyboard Shortcuts

Press Shift+? to see the full shortcuts dialog.

Troubleshooting

Nodes won't connect
If two nodes won't connect, it usually means the consumer node does not support that type of input (e.g., trying to pass a video output into an image-only node). For reference inputs, also check if you have reached the "maximum references" limit for the selected model.

Credit error when running
Use a less powerful model or upgrade your plan.

Node stays in "Queue" or "Processing"
High-quality generations (especially video) can take a few minutes. If a node seems stuck for more than 5 minutes, try refreshing the canvas or re-running that specific node.

Output not showing after running
Ensure your generation node has a prompt. You can either type a prompt directly into the node, connect a Text node to provide dynamic content, or use a source text node.

Group ran but nodes didn't execute in order
Check the connection direction. Flow reads topologically — the upstream node must be fully connected to the downstream node before running.

What's Next

• How to Write Prompts — to get better outputs from your Image and Video nodes
• How to Use Presets — to explore the preset library alongside Flow
• AI Photo Editor — to refine outputs from Flow directly

Ready to build your first pipeline?

Open PhotoGPT Flow →