How do I make my bot generate or edit images with AI?

Use the AI Agent node in Image mode, then send the result with a Message node.

1. Create an image preset

Go to Settings → AI → Presets, create a preset with Preset Type set to Image, and choose the image model. Add an instruction if you want a fixed style.

2. Build the flow

1. A Message node asks: "Describe the image you want, or send a photo to edit."
2. An AI Agent node with Mode: Image, your image preset, and Input Last user message (or Wait for user message).
3. A Message node with an image card whose source is From variable → AI image (last generated).
4. Connect the AI Agent's On error output to a short message, for example "I couldn't create this image, try describing it differently."

How input works

• Text only — draws a new image.
• Photo + text — edits the photo using the text.
• Photo only — edits the photo using the preset's instruction.
• Users can refine step by step ("make it ginger", "add a hat"). The tags #edit and #new let them choose between refining the previous image and drawing a new one. Refining the previous image needs an OpenAI image preset.

Cost and storage

Images use your plan's AI tokens. For example, one medium-quality square image is about 108,000 tokens and a low-quality one is about 32,000. Images made in test mode are free. Generated images are kept for 35 days.

Learn more: AI Agent node, Image mode.