Back to Nodes
Kling V3 Omni Video

Kling V3 Omni Video

Official

Unified multimodal Kling V3 video model for prompt generation, reference-image guidance, and reference-video editing workflows.

Nodespell AI
AI / Video / Kwaivgi

Unified multimodal Kling V3 video model for prompt generation, reference-image guidance, and reference-video editing workflows.

Model Overview

Kling V3 Omni Video combines text-to-video generation with optional image references and video-reference editing pathways. It supports multi-shot JSON prompts and optional native audio generation.

Best At

  • Multimodal generation using text, images, and optional video references.
  • Edit-like workflows that keep continuity from reference video material.
  • Controlled output mode, aspect ratio, and duration tuning.

Limitations / Not Good At

  • More controls increase setup complexity and validation needs.
  • Multi-shot JSON requires careful duration planning.
  • Complex multimodal prompts can require iterative tuning.

Ideal Use Cases

  • Creative video pipelines needing both generation and editing behavior.
  • Scene continuity tasks with reference media guidance.
  • Rapid experimentation across prompt-only and reference-driven flows.

Input & Output Format

  • Input: required prompt; optional start_image, end_image, reference_images, reference_video, mode, video_reference_type, aspect_ratio, duration, generate_audio, keep_original_sound, and multi_prompt.
  • Output: generated video URI returned on response.

Performance Notes

  • pro mode generally targets higher quality with higher cost.
  • video_reference_type changes how reference video is interpreted in generation.
Inputs (6)

Prompt

String

Main text prompt for generation or editing behavior.

RequiredMulti InputMin: 0Max: 100

Multi-Shot Prompt (JSON)

String

Optional JSON shot array, for example [{"prompt":"...", "duration":3}].

Min: 0Max: 100

Start Image

String

Optional first frame reference image.

Min: 0Max: 100

End Image

String

Optional final frame reference image. Requires start image.

Min: 0Max: 100

Reference Images

String

Optional one-to-many reference images for style, subject, or scene guidance.

Multi InputMin: 0Max: 100

Reference Video

String

Optional reference video for style guidance or base-video editing.

Min: 0Max: 100
Parameters (8)

Prompt

String

Main text prompt for generation or editing behavior.

Required
Default:

Mode

String

Generation quality mode.

Default: pro

Video Reference Type

String

Controls whether reference video acts as style guidance or editable base footage.

Default: feature

Aspect Ratio

String

Aspect ratio used when frame or video references do not override framing.

Default: 16:9

Duration

Number

Target video duration in seconds.

Default: 5

Generate Audio

Boolean

Generate native audio with output video.

Default: false

Keep Original Sound

Boolean

Keep sound from reference video when reference video is used.

Default: true

Multi-Shot Prompt (JSON)

String

Optional JSON shot array, for example [{"prompt":"...", "duration":3}].

Default:
Outputs (1)

Output

Inferred

Output

Nodespell Team

Type

Node

Status

Official

Package

Nodespell AI

Category

AI / Video / Kwaivgi

Input

TextImageVideo

Output

Video

Keywords

Video GenerationVideo EditMultimodal GenerationPrompt ConditioningAspect ControlLength Control
Use in Workflow