Nodaro DocsGenerate Script
Write a multi-scene video script with AI. Each scene gets a description, action, mood, length, camera, dialogue and a ready-to-use image prompt.
The Generate Script node writes a structured, multi-scene video script from a short prompt. Each scene comes with a visual description, the action, a mood, a suggested length and an image prompt ready for Generate Image, plus camera directions, dialogue, location details, a music mood and sound effects when they apply. Seven outputs send each part of the script to the node that needs it.
When to use it
- You want a storyboard for a video: one image prompt per scene, then one image and one clip per scene.
- You need a structured script for an explainer, an ad, a product demo or a social video.
- You want dialogue lines, music moods and sound effects planned together with the pictures.
- You want to plan the length and the pacing of a video before you spend credits on generation.
For a complete film made for you from one prompt, use Story → Video.
Quick start
Add the node
Press Tab on the canvas and choose Video › Story & Script › Generate Script.
Describe the video
Add a Text node, type the topic, the story or the concept in it, and connect it to the Prompt input of Generate Script. Be specific: "30-second product ad for a fitness app showing a morning routine" works better than "make a video".
Set the shape of the script
Open the settings panel. Set Number of Scenes and Target Length (seconds), and optionally a Tone and a Style Guide.
Run it and connect the outputs
Click Run. The script appears on the node. Connect its outputs to the nodes that make the pictures, the voices and the music.
Inputs and outputs
Generate Script has five inputs on its left edge:
| Input | Accepts | What it does |
|---|---|---|
| Prompt | Text nodes, such as Text or Generate Text - Any LLM | The topic, the story or the concept. Required: the node has no prompt box of its own. |
| Tone | A Tone node, or any text node | Sets Tone. |
| Style Guide | A Style Guide node, or any text node | Sets Style Guide. |
| Scene Count | A Scene Count node | Sets Number of Scenes, from 1 to 20. |
| Duration | A Duration node | Sets Target Length (seconds), from 5 to 600. |
A connected input wins over the value typed in the panel. While it is connected, the panel shows that field read-only, with a chip that names the connected node. Drag a wire out of one of these inputs to add the matching node.
The seven outputs each carry one part of the active script:
| Output | What it carries | Connect it to |
|---|---|---|
| Scenes | The script as text: the first scene's image prompt. | A text input. |
| Images | The image prompt of every scene. | The Prompt input of Generate Image. With two or more scenes, Generate Image runs once per scene. |
| Dialogue | Every dialogue line of the script, with speaker and emotion. | Text to Dialogue. Click Fill N Lines from Script in its settings to copy the lines. If Text to Dialogue has no lines of its own, it uses the script's lines when it runs. |
| Music | The music moods of all scenes, each once, joined into one text. | The prompt of Generate Music. |
| SFX | The sound effects of all scenes, joined into one text. | The prompt of Text to Audio. |
| Characters | The characters of the script, each once, with name, description, mood, action and position. | Nodes that read script characters. |
| Locations | The locations of the script, each once, with name, description, time of day, weather and lighting. | Nodes that read script locations. |
Settings
| Setting | What it does |
|---|---|
| AI Model | The text model that writes the script. The default is Gemini 3.6 Flash. Search by name, maker or tier. |
| Reasoning Effort | How much the model thinks before it writes, on models that support it. |
| Advanced mode | Gemini and Claude models. Runs the model with the maker's own controls, so that Temperature, Max Tokens and the full reasoning range apply. Those controls appear once it is on. The call is then charged at the rate of the maker's own service. On a Claude model, choosing any Reasoning Effort runs the model the same way, at the same rate. On other models, the switch is disabled with the reason. |
| Number of Scenes | How many scenes to write, from 1 to 20. The default is 5. The Scene Count input can set it. |
| Style Guide | Visual and narrative directions for every scene, such as a color palette, an era or a visual reference. It is sent to the model with the topic, and it can name other nodes as {Node Label}. The Style Guide input can set it. |
| Tone | The emotional register of the whole script, for example whimsical, dramatic or educational, up to 200 characters. The Tone input can set it from a Tone node or any text node. |
| Target Length (seconds) | The total length the script plans for, from 5 to 600 seconds. The default is 60. The AI spreads it across the scenes, so single scenes can vary. The Duration input can set it. |
| Pre & post text | Text added before and after the prompt, not the style guide, when the node runs. It is hidden from people who use your workflow as an app. See Prompt pre and post text. |
To keep these values in one place for several scripts, connect Generation Settings nodes to the matching inputs instead of typing them.
What each scene contains
| Part | What it is |
|---|---|
| Scene number and name | The position and a short name of the scene. |
| Visual description | A detailed description of what is seen. |
| Action | What happens in the scene. |
| Mood | One or more emotional tones. |
| Duration | The suggested length in seconds. |
| Image prompt | A prompt ready for Generate Image. |
| Characters | Each character's name, description, mood, action and position. |
| Dialogue | Spoken lines, each with a speaker, the text and an emotion. |
| Location | The name, a description, the time of day, the weather and the lighting. |
| Cinematography | The shot type, the camera angle and the camera movement. |
| Music mood and sound effects | Suggestions for the soundtrack. |
Edit the script
After a run, the settings panel shows the generated script: its Title, the number of scenes and the total length, and one section per scene. Open a scene to edit its Visual Description, Action, Mood, Duration (s) and Image Prompt (for Generate Image). Your edits are what the outputs send. Copy Prompts copies the image prompts of all scenes.
Every run is kept in the node's result strip, and the outputs follow the result you select.
Once the script exists, Expand Storyboard on the node creates the workflow nodes for every scene. Choose Scene Nodes, which creates one scene node per scene, or Pipeline Nodes, which creates separate Generate Image, video, Text to Speech and Merge Video & Audio nodes per scene. You can add a Combine Videos node at the end and start the image generation right away. The dialog shows the estimated credits before you create the nodes.
Presets
Open Presets on the node for format presets that set the tone, the number of scenes and the target length. Type your topic in the prompt.
| Folder | Presets |
|---|---|
| By Format | YouTube Short, Explainer, Ad Spot, Product Demo, Listicle, UGC Ad (a selfie-style Hook, Show, Proof and Opinion, with figures and proof taken only from your topic) |
| Long-Form & Narrative | Podcast Outline, Trailer Narration, Story Beats |
Credits
A run is one call to the AI model. The call is charged for the tokens it actually used, at the model's rate. The tokens are the topic, the style guide and tone, and the script the model writes.
- Before the run, the Run button shows an estimate (
≈and a number) of a typical script on the selected model. The default, Gemini 3.6 Flash, costs less than Claude Sonnet 5.5. - A longer script costs more. More scenes or a longer target length write more text.
- A hold comes first. When the run starts, Nodaro holds enough credits for the longest answer the model may write. After the call, the run is charged what it used.
- Reasoning Effort has no price of its own. Advanced mode, and any reasoning effort on a Claude model, run the call on the maker's own service, at that service's rate. The cost badge on the node updates when you turn Advanced mode on.
- Failures. A script that the model's maker billed but that could not be read back as scenes is charged for that usage. A failure that was not billed is refunded.
The full rules are in How LLM calls are charged. See Choosing a model to compare models.
Tips
- Be specific in the prompt. Name the product, the audience, the length and the format.
- Use Tone for consistency. It applies to every scene, so the whole script keeps one emotional register.
- Put the look in the Style Guide. A palette, an era or a visual reference keeps the image prompts consistent.
- Treat the length as a plan. Target Length guides the AI, but each scene's length can vary with its content.
Troubleshooting
The run stops with "needs a topic" or "no prompt found". The script has nothing to write about. Connect a Text node with the topic, the story or the concept to the Prompt input. The Tone, Style Guide, Scene Count and Duration inputs set settings only; they are not a topic. The run stops before it starts, so nothing is charged.
From the API
POST /v1/generate-script writes a script from code. It answers at once with a jobId. The finished script is in the job's output_data.script, with title, totalDuration and scenes. See Jobs.
| Field | What it does |
|---|---|
prompt | Required. The topic, the story or the concept, up to 100,000 characters. |
sceneCount | Number of Scenes, from 1 to 20. The default is 5. |
targetDuration | Target Length (seconds), from 5 to 600. The default is 60. |
tone | Tone, up to 200 characters. |
styleGuide | Style Guide, up to 100,000 characters. |
llmModel | The text model. The default is gemini-3.6-flash. |
reasoningEffort | none, low, medium, high, xhigh or max. |
advancedMode, temperature, maxTokens | Advanced mode and its controls, on Gemini and Claude models. |
The MCP tool is generate_script, which takes the style guide as style_guide. See also Run a single node.
Frequently asked questions
What does the Generate Script node produce?
A structured script with a title, a total length and a list of scenes. Each scene has a visual description, the action, a mood, a suggested length, an image prompt, and when relevant the characters, dialogue, location, camera, music mood and sound effects.
How do I make one image per scene?
Connect the Images output of Generate Script to the Prompt input of a Generate Image node. When the script has two or more scenes, Generate Image runs once per scene, with that scene's image prompt.
How many credits does Generate Script cost?
A run is one call to the AI model, charged for the tokens it used at the model's rate. The Run button shows an estimate of a typical script before you run. More scenes or a longer target length write more text and cost more.
How many scenes should I ask for?
About one scene for every 5 to 10 seconds of video works well for most formats. Number of Scenes accepts 1 to 20, and Target Length sets the total length the AI plans for.
Related
Generate Image
Generate Video
Text to Dialogue
Story → Video
Generate Music
Last updated on
Content Ideas
Turn one or more content recipes and your brand into 1 to 10 post ideas, each with a hook, beats and a shot list. The next node runs once per idea.
AI Avatar
Make a talking avatar video with HeyGen. Pick an avatar or animate your own image, then give it a script and a voice, or a recorded voiceover.