> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orisu.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Generate Text

> Generate text from a prompt with an LLM, optionally with media attachments and brand context.

Generate Text uses an LLM to produce text from a prompt. You can attach media for the model to read and supply brand context to keep the output on-brand.

This node runs on a selectable model and consumes credits when it runs.

## When to use

* Writing copy from a brief
* Summarizing documents
* Generating product descriptions

## Creative assistants

Creative assistants are pre-filled starting points. Each one sets an expert system prompt, a suggested model, and sensible defaults so you can skip the prompt engineering, then edit the instructions however you like.

Pick one in two places:

* In the **node picker**, hover **Generate Text** to reveal the assistants grouped by category, then click or drag one onto the canvas. The new node is named after the assistant.
* In the **config panel**, open a Generate Text node and choose an assistant under **Creative assistant**.

Assistants are grouped by what they help you make:

| Category           | Examples                                                                 |
| ------------------ | ------------------------------------------------------------------------ |
| Image prompting    | Nano Banana prompter, Enhance image prompt, Static ad prompter           |
| Video prompting    | Kling cinematic prompter, Sora prompter, Storyboard planner              |
| Audio & voice      | ElevenLabs dialogue prompter, Music prompter, Sound effects prompter     |
| Brand intelligence | Extract brand guidelines, Extract brand positioning, Extract brand voice |
| Copywriting        | Product description writer, Ad copy & hooks                              |

Selecting an assistant only pre-fills the node. Your edits to the instructions always win, and you can clear the assistant at any time to start from a blank prompt.

## Inputs

<ParamField body="Text Prompt" type="text">
  The prompt to generate from. Optional. Accepts an array and supports loop mode.
</ParamField>

<ParamField body="Context" type="any">
  Anything to include as context: images, videos, files, scraper Results, text, or mixed bundles. Images are attached natively for vision-capable models. Each item is described to the model with a stable reference name (`[post_1]`, `[image_2]`, `[text_1]`) so you can refer to specific items in your prompt. Optional, accepts multiple inputs and an array, and supports loop mode.
</ParamField>

<ParamField body="Brand" type="brand">
  Brand context to guide the output. Optional, accepts multiple inputs and an array.
</ParamField>

<ParamField body="Product" type="product">
  Product context from a [Product from Website](/nodes/integrations/extract-product) node. The product's name, description, features, price, and audience are injected into the prompt so the copy stays on-product. Optional, accepts multiple.
</ParamField>

At least one of Text Prompt or Media must be connected, or the node has no input to process.

## Outputs

<ParamField body="Text" type="text">
  The generated text.
</ParamField>

## Referencing connected inputs

Type `@` anywhere in the Instructions field to open an autocomplete list of the node's connected inputs. Selecting one inserts a live chip. Rename the source node and the chip updates; disconnect the edge and the chip turns grey to show it is unresolvable.

At run time the chip resolves to a stable label for media inputs (e.g. `[image_1]`, matching the legend the node already attaches) or the inlined text value for text inputs. This means you can write instructions like:

> "Use the product shown in `@Image` as the hero and keep the tone from `@Text Prompt`."

and the model reliably receives the right content for each reference.

<Note>
  A reference points at a connected source as a whole. If one connection carries several media items, the reference resolves to that source's first item (e.g. `[image_1]`); every item still appears in the attached legend.
</Note>

## Configuration

<ParamField body="instructions" type="string" default="">
  System instructions for the model. Pre-filled when you pick a creative assistant; editable at any time.
</ParamField>

<ParamField body="assistantPreset" type="string" default="null">
  The creative assistant the node started from, if any. Set when you pick an assistant and used to show which one is active.
</ParamField>

<Note>
  This node runs on a selectable LLM. See [Text models](/models/text) for supported families and [Model settings](/models/settings) for options like reasoning effort and web search.
</Note>