Creative ToolsAudio and text tools
Write or transform text with Generate Text
Generate Text is the Nodespell Creative Tool for creating or transforming text with a language model. You write a prompt—and can add images for the model to look at—and get back text: a script, a set of prompts, a caption, a description of a picture. The result is a text output you can read, copy or connect to other nodes.
When to use Generate Text
Section titled “When to use Generate Text”| You want… | Use |
|---|---|
| A script, shot list, caption or rewritten prompt | Generate Text |
| A description of an image | Generate Text with the image added |
| Help building or changing the graph, or advice about your project | Agent |
| Spoken audio of a script | Voice Generator |
Generate Text runs one model on one prompt and returns its answer. It does not read your graph or make changes to it; the Agent does that.
Generate text
Section titled “Generate text”- Switch the workspace to Create and open All tools, then select Generate Text. Generate Text has no button of its own in the tool rail.
- Check the model. The default is GPT Text.
- Write the prompt. To have the model look at pictures, add them to the images field.
- Optionally change System Instructions to give the model a role or rules, such as “You write short, visual video prompts.”
- Set Generations from 1 to 4 if you want several versions.
- Select Generate.
Generate Text controls
Section titled “Generate Text controls”| Control | What it does | Options and defaults |
|---|---|---|
| Model picker | Chooses the text model | Default: GPT Text |
| Previous model / Next model | Steps through text models | — |
| More settings | Less-used model settings, with a count of changed values | Collapsed |
| Generations | Number of runs submitted together | 1 to 4, default 1 |
| Generate | Adds or updates the model node and submits the runs | Shows the credit estimate |
Recommended models for Generate Text
Section titled “Recommended models for Generate Text”The Featured collection lists GPT Text (default), Claude Text and Gemini Text. All models also lists Deepseek R1 and Deepseek V3. Each featured model has a Model setting that picks a version within that family:
| Model | Model options | Default |
|---|---|---|
| GPT Text | GPT 5 Nano, GPT 5 Mini, GPT 5 | GPT 5 Nano |
| Claude Text | Claude 3.5 Haiku, Claude 4.5 Sonnet, Claude 4.1 Opus | Claude 3.5 Haiku |
| Gemini Text | Gemini 2.5 Flash Lite, Gemini 2.5 Flash, Gemini 2.5 Pro | Gemini 2.5 Flash Lite |
All three featured models share these settings:
| Setting | What it does | Default |
|---|---|---|
| Prompt | The request or text to process | — |
| System Instructions | The persona or rules the model follows | “You are a helpful assistant.” |
| Maximum output tokens | Caps the length of the answer, up to 3,000 tokens | 1000 |
| Images input | Pictures the model can read | Optional |
The output-length setting is labelled Max Output Tokens on GPT Text and Gemini Text, and Maximum Output Tokens on Claude Text. See Choose a model for switching.
Use the text in your graph
Section titled “Use the text in your graph”Generate Text places a text model node in the current graph level. In Graph you can connect its text output to the prompt input of an image, video or audio node, so the generated text drives the next step. When a prompt is fed this way, the Create form shows “Prompt controlled by an input node” on the receiving tool. See Connections and Prompts and saved text.
Where does the text go?
Section titled “Where does the text go?”The answer appears in Results in the Create form and stays in the node’s history; see Takes. To read a long answer, open it in the Design panel, which shows text with search, wrapping and Markdown views; see Viewers.
Common problems
Section titled “Common problems”- The answer stops mid-sentence. Raise the output-token setting, up to 3,000.
- The answer ignores the image. Make sure the image is in the images field, and ask about it directly in the prompt.
- Answers are too generic. Put the format and constraints in System Instructions, and give an example in the prompt.
Frequently asked questions
Section titled “Frequently asked questions”What is the default Generate Text model?
Section titled “What is the default Generate Text model?”The default is GPT Text, using GPT 5 Nano unless you change its Model setting. Claude Text and Gemini Text are also featured.
Can Generate Text describe an image?
Section titled “Can Generate Text describe an image?”Yes. GPT Text, Claude Text and Gemini Text accept images; add them to the images field and ask about them in the prompt.
How long can a Generate Text answer be?
Section titled “How long can a Generate Text answer be?”GPT Text, Claude Text and Gemini Text allow up to 3,000 output tokens, with a default of 1,000.
Is Generate Text the same as the Agent?
Section titled “Is Generate Text the same as the Agent?”No. Generate Text runs one model on your prompt and returns text. The Agent can inspect your project and propose or make graph changes.
Can I use generated text as a prompt for an image or video?
Section titled “Can I use generated text as a prompt for an image or video?”Yes. Connect the Generate Text node’s output to the prompt input of another node in Graph.
Related
Section titled “Related”- Agent — an assistant that works with your graph
- Prompts and saved text — prompt nodes and reusable text
- Connections — feed text into other nodes
- Data and logic nodes — process text further
- Choose a model — compare text models