Text Generate

Text Generate is the text node. It holds a conversation with a language model the way a chat app does: you write a message, the reply streams in, and the node remembers the exchange. Each node keeps its own memory, so two Text Generate nodes on one canvas are two separate conversations.

Use it for the words around a production: captions and product copy, a prompt draft you refine before it goes to an image or video model, a list of variations, or a question about a picture or a document you attach.

1. 1

   Add Text Generate

   Find Text Generate in the Generative group and add it to the canvas.
2. 2

   Choose a model

   Open the compact model menu at the bottom of the field. Models are grouped by provider and show their cost per message.
3. 3

   Attach what it should read

   Press the attach button to add pictures or files, or wire them in. Attachments ride with the next message only.
4. 4

   Write and send

   Type the message and press Send. The reply streams into the node as it is written.
5. 5

   Keep going

   Follow up in the same conversation. The model sees the whole exchange, so you can ask for shorter, bolder, or a different angle.

The attach button and the Attachments port feed the same strip, shown as thumbnails inside the field. It takes pictures, text from a Text Input, and files from a File Upload node: PDF, plain text, Markdown or JSON. The model reads them with the message you send next.

After that message the strip clears itself. Attachments are not re-sent, and not paid for again, on every later turn. To ask a second question about the same file, attach it again. Each model sets how many attachments one message can carry.

Model

Chooses the language model. It locks after the first reply, because switching mid-conversation would replay the whole exchange on a different model.

Message field

Where you write. Attachments pop in as thumbnails above the text.

Send

Sends the message and shows the credit cost of one reply for the selected model.

Copy reply

Copies a reply to the clipboard, for use outside the canvas.

Extract

Creates a Text Input node beside this one, pre-filled with that paragraph, ready to wire into a prompt or a variable.

Clear Conversation

Empties the memory and unlocks the model menu. Use it to start a fresh topic without adding a node.

Extract is the bridge from the chat to the graph. Press it on any paragraph of a reply and a Text Input appears next to the node with that text in it. Wire the Text Input into an Image Generate or Video Generate prompt, or into a prompt variable such as [Headline], and the words the model wrote drive the next generation.

Extract as many paragraphs as you like. A reply that lists five directions can become five Text Inputs feeding five generators, while the conversation that produced them stays intact for the next round.

A message is priced per reply at a flat rate that depends on the model, shown on the Send button before you press it. Attachments sent with the message are included in that reply. Every exchange is recorded as one step, so Undo takes back a whole message and its reply together.

• Image Preview → Attachments

  Ask about a result: describe it, critique it, or write the caption for it.

• File Upload → Attachments

  Give the model a brief, a brand guide or a product sheet to write from.

• Text Input → Attachments

  Hand over a prompt or copy you already have, for the model to rewrite or extend.

• Extract → Text Input → Image Generate

  Turn a drafted prompt into the generation it describes.

• Extract → Text Input → prompt variable

  Feed a headline, a product name or a line of copy into a fixed prompt.