Text nodes
Text nodes
Write prompts by hand or let AI craft them for you. Text is where most creative workflows begin.
Text, or Prompt Generator, is the universal connector in Spaces. Most generator nodes (image, video, audio) accept a text prompt as input, which makes text nodes the natural starting point for just about everything you build. A single prompt can feed an Image Generator and a Video Generator at the same time. Or an Assistant can refine a rough idea into a polished prompt before passing it along.
In this article
Text
The Text node is the simplest and most versatile node in Spaces. It holds static text content that you can connect to any node accepting text input. Write once, connect to many.
How to use it
Add a Text node from Spotlight or the node panel. Type your prompt or content in the text area on the card. The content flows through the Text output port to any connected downstream node.
The text area supports rich formatting and @mentions. Type @ followed by a node name to dynamically reference another node's output in your text. The referenced content updates automatically when the source changes.
Formatting options
When you click the node or start typing, a toolbar appears with formatting controls: text style (Paragraph, Heading 1, 2, 3), bold, italics, bullet and numbered lists, and horizontal rules. You can also change the node background color for visual organization.
Input and output
The Text node is a source node. It has no input ports. It produces a single Text output that you connect to generators, the Assistant, or any node that accepts text.
Use cases
- Simple prompt source. Write 'A futuristic city at sunset' and connect it to an Image Generator.
- Shared prompt. One Text node connected to both an Image Generator and a Video Generator. Same prompt, different media.
- Dynamic references. Use @mentions to reference an Assistant's output. The Assistant writes the prompt, the generator uses it, and it all stays in sync.
Assistant
The Assistant uses AI large language models to generate, refine, or transform text. Give it instructions and it produces intelligent output. It is perfect for crafting optimized prompts, writing scripts, analyzing images, or any text task where you want AI help.
How to use it
Add the node
Search for Assistant in Spotlight. You can also search for a specific model like Claude or Gemini to add a pre-configured node.
Write your instructions
On the Prompt tab, describe what you want the AI to do. You can use @mentions to reference other nodes.
Choose a model
Click the model selector to pick from OpenAI, Google, Anthropic, or xAI models.
Connect inputs (optional)
Connect images to the Media port for the AI to analyze, text to the Texts port, or video to the Video port.
Set the output mode
Choose Simple response for a single text output, or As list for structured list output (useful for batch workflows).
Run the node
The AI-generated text appears on the Result tab, ready to flow to connected nodes.
Settings
| Setting | What it does |
|---|---|
| Instructions | Your prompt for the AI. Supports @mentions to reference other nodes. |
| Model | Which AI model to use (see table below). |
| Output Mode | Simple response (single text) or As list (structured list). |
| System prompt | Optional instructions that shape the AI's behavior and personality. |
Input and output
| Direction | Port | Data type | Notes |
|---|---|---|---|
| Input | Media | Image (multiple) | Images for the AI to analyze or describe |
| Input | Texts | Text (multiple) | Text for the AI to process or incorporate |
| Input | Video | Video (multiple) | Video for the AI to analyze |
| Output | Generated Text | Text | The AI-generated response |
Use cases
- Prompt engineering. Write rough instructions, let the Assistant refine them into an optimized image prompt, then connect that to an Image Generator.
- Content analysis. Connect an image to the Media port. The Assistant describes it in detail. Use that description as a prompt for Variations or further generation.
- Script writing. Give instructions for a voiceover script. The Assistant writes it. Connect the output to a Voiceover node.
- Batch processing. Connect a List of topics. The Assistant generates a prompt for each one. An Image Generator creates images for all of them.
Assistant models
The Assistant supports models from four providers. Each can be selected directly from Spotlight as a variant (for example, search "Claude" to add an Assistant pre-configured with that model).
| Model | Provider | Best for |
|---|---|---|
| GPT-5 Mini | OpenAI | Fast, versatile responses |
| GPT-4.1 Mini | OpenAI | Balanced speed and quality |
| GPT-5.2 | OpenAI | High-quality reasoning |
| Gemini 3 Pro | Advanced multimodal analysis | |
| Gemini 3 Flash | Fast multimodal processing | |
| Gemini 2.5 | Balanced Google model | |
| Gemini 2.5 Lite | Lightweight, fast | |
| Grok 4.1 Fast | xAI | Fast, creative responses |
| Claude Sonnet 4.5 | Anthropic | Nuanced, thoughtful writing |
| Claude Opus 4.5 | Anthropic | Highest quality reasoning |
Tips and best practices
One Text node, many generators. Connect a single Text node to both an Image Generator and a Video Generator to create different media from the same prompt.
Use @mentions for dynamic prompts. Reference other nodes' outputs in your text. When the source changes, your prompt updates automatically.
Color-code your Text nodes. Use background colors to visually distinguish different types of prompts on your canvas (for example, blue for image prompts, green for video prompts).
Set a System Prompt on the Assistant. It shapes the AI's personality and focus. For example, "You are a professional prompt engineer specializing in photorealistic image descriptions" produces very different results than a generic setup.
Connect images to the Assistant. The AI can see and describe images connected to the Media port. This is great for generating accurate descriptions or analyzing visual content.
Use "As list" mode for batch workflows. When the Assistant outputs a list, it connects naturally to List nodes for processing each item through the rest of your workflow.
Match the model to the task. GPT models are strong for general tasks, Gemini excels at multimodal analysis (images + text), and Claude is great for nuanced creative writing.