An LLM node that writes, sharpens or extracts text — pick Generate, Improve or Extract mode, choose an output format, and feed it any upstream media as typed context.
Where: canvas → left rail → Add → Words & data → Gen Text / Prompt — Add a Text Generation node and set a mode
Key ideas
- Three modes: Generate writes formatted text from your prompt + context; Improve rewrites a prompt into a sharper version; Extract reads connected media/docs/code and produces a usable prompt.
- Output format: Choose Markdown, Plain, HTML, JSON or List for the output shape.
- Unified context input: The single Context (ctx_in) handle accepts any source — images, video, audio, SVG, HTML, PDF, 3D, design systems, research summaries — collected into one typed list the LLM reads. Tone and length hints can also be set.
- Dual outputs: The node exposes a plain Text output and an HTML/MD output so you can pipe prose downstream as either.
- Gen Text / Prompt card: TextGenerationNode uses the authored shared-card path. Generate versioned text/prompt output with model, system prompt, formatting, and connected context.
- Properties sidebar: genOutputNodes.tsx
- Opened surfaces: Rich text preview/editor; AI co-writer; import/export
Steps
- Add a Text Generation node.
- Pick a mode (Generate / Improve / Extract) and an output format (MD / Plain / HTML / JSON / List).
- Type a prompt or topic into the Prompt handle and wire any media/docs into the Context handle.
- Click Generate.
- Copy the result, or wire text_out / html_out into another node (e.g. a Video prompt or Section).
Gen Text / Prompt controls and presets
- Add path: Flowgen → Add → Inputs → Gen Text / Prompt.
- Card: TextGenerationNode.
- Properties: genOutputNodes.tsx.
- Controls: Prompt/system prompt; model/thinking; context/references; formatting; versions.
- Opened surfaces: Rich text preview/editor; AI co-writer; import/export.
Connected rich context
Gen Text can include compatible connected images, video, SVG, PDF/document material, and 3D/media descriptors alongside text context. The chosen model determines which attachments it can actually consume.
Troubleshooting
A connected file is ignored
Confirm it reaches a context/reference input and choose a text model that supports that attachment kind/size.