These nodes cover text creation, combination, and prompt generation inside Workflows. Use them whenever you need to control or automate what your prompts say before they reach a generator.

The Text node sends simple text into other nodes, primarily Prompt Template, Video Timeline, Concatenate Text, and Prompt Generator nodes. Wherever text can be used, a Text node lets you hardcode it or provide it as a workflow input.

A Text node can feed into an Edit or Create node as a prompt, but it cannot feed into nodes that specifically require text, like Concatenate Text or the text inputs of the Video Timeline node. The Text node is a basic node with no other parameters beyond the text itself.

The Concatenate Text node combines different text nodes into new combinations. For example, if one text group has "Fish" and "Sharks" and another has "are cute" and "are scary," the node produces "Fish are cute," "Fish are scary," "Sharks are cute," and "Sharks are scary." Text is joined using a separator you choose, most commonly a space, though any text works, including full sentences or paragraphs.

The Prompt Template node turns a prompt into a template you can partially fill in at runtime. Type "@" followed by a label to create a template area, then link another node's output into that area. This lets a workflow require just the small portion of a prompt that should vary, rather than the whole prompt.

Example: "Show an @animal eating a @food." The @animal reference is replaced with whatever text is attached to the "animal" input, and @food is replaced the same way.

Multiple nodes can feed the same input and combine in the result. For example, a "Mammals" text node and a "Reptiles" text node can both feed into @animal, giving you both types in the output. You can also create multiple templates that reuse the same inputs, mixing and matching them across templates.

The Prompt Generator node sends a prompt to an LLM and gets text back, generating up to 25 distinct prompt variations from a single instruction. Be as specific as possible for best results. Asking for "a cute cat" produces inconsistent variety, while "a cute cat sitting down, in a modern flat icon style like those used for SVG or iconography" produces much more usable, consistent results.

You can also use this node to choose between lists, return specific words, or generate other short text, which is useful when the output feeds into a Prompt Template node later in the workflow.

The Prompt Generator node is not a substitute for an agent. It doesn't retain conversational context, and its output can't be edited within the graph; treat each request as a one off text generation. Also note that mentioning a number of responses in the prompt text itself doesn't change how many you get back. Use the node's slider to set that instead.