Scenes

Scenes are the fundamental video generation nodes in Dal Nulla. Each Scene Node produces a video clip based on your text prompt and connected reference images. Scenes are where your creative vision comes to life — from AI-generated video clips to video-to-video transformations.

What is a Scene Node?

A Scene Node represents a single video clip in your project. It takes text prompts, reference images, start/end frames, or even other videos as input, and produces a generated video as output. Scene Nodes are the core of any video workflow in Dal Nulla.

You can chain multiple Scene Nodes together to build sequences, use video-to-video (V2V) to transform existing clips, or connect reference images to guide the AI’s visual output.

Inputs & Outputs

Port	Direction	Type	Description
input	In	Text	Text prompt from Text Node, Concatenator, List Selector, or JSON Parser
source	In	Image	Reference images (up to 3) from Reference Image, Draw, Upscaler, or Network Image nodes
startFrame	In	Image	Start frame image for Frames mode
endFrame	In	Image	End frame image for Frames mode
sourceVideo	In	Video	Source video from another Scene for video-to-video (V2V)
output	Out	Video	Generated video to other Scenes (V2V), Upscaler, AI Voice, Text (multimodal), Image, Social nodes, AI If/Else, or Google Drive Export

Inspector Controls

When you select a Scene Node, the Inspector Panel on the right lets you configure every aspect of the video generation.

Prompt

Describe the action, style, and content of the video. If a Text Node is connected, its content appears here automatically. You can Detach it to edit the prompt independently, or Expand the editor for a full-screen writing experience.

Negative Prompt

Specify what to avoid in the generated video. Common values: blur, distortion, text, watermark, low quality, static.

Model

Choose the video generation model. Each model has different strengths, quality levels, and pricing. See the full model table below.

Duration

Set the video length. Options depend on the selected model — typically ranging from 3 seconds to 15 seconds, with VEO 3.1 supporting up to 148 seconds.

Resolution

Choose the output quality: 720p, 1080p, or 4K. Note that 4K availability is model-dependent and more expensive.

Aspect Ratio

Set the video dimensions: 16:9 (landscape), 9:16 (portrait/vertical), 1:1 (square), 4:3, 3:4, and more depending on the model.

Audio

Toggle audio generation on or off. When ON, the AI generates matching audio for the video. When OFF, the video is silent — and costs fewer credits.

Seed

Set a specific seed number for reproducible results. Using the same seed with the same prompt and settings produces consistent output. Supported by most models.

Media Input Modes

Toggle between two ways of controlling visual input:

Refs (References): Connect up to 3 Reference Images. The AI blends them for style and content guidance.
Frames: Set a Start Frame and/or End Frame. The AI animates the transition between these images. Great for product reveals and transitions.

Use Global Context

Include the project-wide global context (text and reference images) in this scene’s generation. Uncheck if you want the scene to be completely independent.

Version History

Every generation is saved in the history. Browse previous versions and click to restore any prior video and prompt combination.

Tag

Assign a tag (e.g., @scene-1) to this node so you can reference its output in downstream Text Nodes for multimodal analysis.

Available Models

Model	Provider	Tier	Max Res	Durations	FPS	Img2Vid	V2V	Best For
VEO 3.1 Pro	Google	Free	4K	8-148s	-	Yes	No	Long clips, highest quality, free
VEO 3.1 Fast	Google	Free	1080p	8-148s	-	Yes	No	Quick generation, free
Grok Imagine Video	xAI	Pro	720p	5/10/15s	-	No	Yes	V2V transformations
Kling 3.0 Pro	KlingAI	Pro	1080p	3-15s	-	Yes	No	High quality, flexible duration
Kling 3.0 Standard	KlingAI	Pro	1080p	3-15s	-	Yes	No	Good balance of speed and quality
Kling O3 Pro	KlingAI	Pro	1080p	3-15s	-	Yes	No	Advanced KlingAI
Kling O3 Standard	KlingAI	Pro	1080p	3-15s	-	Yes	No	Standard KlingAI
Kling 2.6 Pro	KlingAI	Pro	1080p	5-10s	-	Yes	No	Reliable quality
Sora 2 Pro	OpenAI	Pro	1080p	4/8/12s	-	Yes	No	Cinematic quality
Sora 2	OpenAI	Pro	720p	4/8/12s	-	Yes	No	Fast OpenAI generation
Runway Gen-4.5	Runway	Pro	1080p	5/8/10s	24	Yes	No	Consistent characters
Runway Gen-4 Turbo	Runway	Pro	720p	2/5/10s	-	Yes (required)	No	Fast, requires start image
Runway Aleph	Runway	Pro	1080p	5/10s	-	No	Yes (required)	V2V only
Hailuo 02	MiniMax	Pro	1080p	6s	-	Yes	No	High quality short clips
Hailuo 2.3	MiniMax	Pro	1080p	6s	-	Yes	No	Latest Hailuo
Hailuo 2.3 Fast	MiniMax	Pro	1080p	6s	-	Yes (required)	No	Fast, requires image
Seedance 1.5 Pro	ByteDance	Pro	1080p	4-12s	24	Yes	No	Smooth motion
Seedance 1.0 Pro	ByteDance	Pro	1080p	5-10s	24	Yes	No	Reliable quality
Seedance 1.0 Pro Fast	ByteDance	Pro	1080p	5-10s	24	Yes	No	Quick Seedance
Seedance 1.0 Lite	ByteDance	Pro	720p	5-10s	24	Yes	No	Budget Seedance

Tier: “Free” models are available on all plans. “Pro” models require an active subscription.

How to Use

Add a Scene Node to the canvas — right-click and select “Scene”, or drag it from the sidebar.
Write a descriptive prompt for the video clip (or connect a Text Node).
Select a video model based on your quality, speed, and cost needs.
Choose duration, resolution, and aspect ratio from the inspector controls.
(Optional) Connect Reference Images in Refs mode for visual guidance, or use Frames mode with a Start Frame and End Frame.
(Optional) Toggle audio on or off.
Click Generate Scene or run the full workflow.

Workflow Examples

Multi-Reference Video

Connect a Text Node with “cinematic scene of a warrior walking through a forest” to a Scene Node. Connect two Reference Images — one of the character (tag: @warrior) and one of the forest background (tag: @forest) — to the same Scene Node’s source ports. The AI generates a video blending the character with the forest setting.

Video-to-Video Chain

Scene A generates an initial clip. Connect Scene A’s output to Scene B’s sourceVideo port. Scene B applies a V2V transformation (e.g., style transfer) while preserving the original motion. Then connect Scene B to an Upscaler Node to enhance the final quality.

Sequential Story

Create a sequence of scenes that flow together. Connect Text Nodes with chapter prompts to each Scene. Use shared Reference Images across scenes to maintain character consistency throughout the story.

Tips & Best Practices

Free models (VEO 3.1) support up to 148 seconds and 4K resolution — great for both prototyping and production.
Use Frames mode for precise start/end control. This is ideal for product reveals and smooth transitions.
V2V (Video-to-Video) transforms existing video while preserving motion — only Grok Imagine Video and Runway Aleph support this.
Turn audio OFF for silent clips to reduce generation cost.
Connect up to 3 reference images in Refs mode for maximum style control.
Use tags (@scene-1) to reference video output in downstream Text Nodes for multimodal analysis.
4K is expensive — prototype at 720p or 1080p, then regenerate your best scenes at 4K.
Use the seed parameter when you want consistent results across iterations.
Lock a scene node to preserve its generated video during workflow re-runs.

Troubleshooting

Problem	Cause	Solution
Long generation times	Video generation is inherently slow (30s-5min)	Use Fast models (VEO 3.1 Fast, Hailuo 2.3 Fast, Seedance Fast) for quick iterations
Reference images ignored	Model does not support img2vid	Check the model table above. Note that Runway Gen-4 Turbo requires a start image
V2V not working	Model does not support V2V	Only Grok Imagine Video and Runway Aleph support V2V. Connect the source video to the `sourceVideo` port
4K unavailable	4K is model-dependent	VEO 3.1 Pro and Kling models support 4K. Other models max out at 1080p
Audio quality issues	AI-generated audio is experimental	For professional narration, use AI Voice Nodes instead
Video looks static	Prompt lacks motion verbs	Use action verbs like “running”, “flying”, “panning”, “zooming” in your prompt