Back to Nodes
Cosmos 3 Super Image to Video

Cosmos 3 Super Image to Video

Official

Animate a first-frame image with NVIDIA Cosmos 3 Super, using prompt expansion and optional agentic refinement.

Nodespell AI
AI / Video / Nvidia

Animate a first-frame image with NVIDIA Cosmos 3 Super, using prompt expansion and optional agentic refinement.

Use this for image-to-video motion tests where a source frame anchors the scene and the prompt describes movement, camera behavior, or action.

Inputs (2)

Image

String

First-frame image for the video.

RequiredMin: 0Max: 100

Prompt

String

Prompt describing the motion and scene.

RequiredMulti InputMin: 0Max: 100
Parameters (10)

Prompt

String

Prompt describing the motion and scene.

Required
Default:

Frames

Number

Number of frames to generate.

Default: 189

FPS

Number

Frames per second of the output video.

Default: 24

Guidance Scale

Number

How strongly the result follows the prompt.

Default: 6

Inference Steps

Number

Number of denoising steps.

Default: 28

Prompt Expansion

Boolean

Rewrite the prompt into a denser caption before generation.

Default: true

Agentic Refinement

Boolean

Generate and critique multiple candidates through an iterative refinement loop. Costs scale with candidates.

Default: false

Agentic Iterations

Number

Maximum prompt refinement stages when agentic refinement is enabled.

Default: 2

Agentic Samples

Number

Candidate videos generated and judged per agentic iteration.

Default: 2

Seed

Number

Optional seed for repeatable generation. Set to -1 for a random seed.

Outputs (1)

Output

Inferred

Generated video output.

Nodespell Team

Creator profile

Type

Node

Status

Official

Package

Nodespell AI

Category

AI / Video / Nvidia

Input

ImageText

Output

Video

Keywords

Video GenerationImage To VideoPrompt ConditioningLength Control
Use in Workflow