# Kling v3 Omni Video - Core.Today AI API

> Kling Video 3.0 Omni: a unified multimodal video model that generates and edits video from text, images, reference images, and existing video. Combines text-to-video, image-to-video, reference-based generation, and video editing with native audio and multi-shot control.

- **Provider**: Kuaishou
- **Model ID**: kwaivgi/kling-v3-omni-video
- **Category**: Video Generation
- **Credits**: 2610 per 5s pro (1080p) video, no audio (default) — billed per second: standard 391-521/s, pro 521-652/s, 4k 978/s
- **Speed**: Slow
- **Quality**: Ultra

## Features
- Three quality tiers: standard (720p), pro (1080p, default), and 4k
- Reference-based generation from up to 7 images for character/style consistency
- Video editing mode via reference_video + video_reference_type=base
- Style and camera reference via reference_video + video_reference_type=feature
- Multi-shot control via multi_prompt (up to 6 sequential shots)
- Optional native audio generation (off by default)

## Use Cases
- Reference-image-driven brand or character-consistent video series
- Editing an existing video clip with text instructions
- Multi-shot mini-trailers with a single coherent prompt
- 4K hero shots for premium marketing content
- Style transfer from a reference video's camera work

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Required
- **prompt**: string - Text prompt (max 2500 chars). Supports <<<image_1>>>, <<<video_1>>> template references

### Optional
- **start_image**: string - First frame image for image-to-video
- **end_image**: string - Last frame image (requires start_image)
- **reference_images**: array - Reference images for elements, scenes, or styles (max 7 without video, 4 with video)
- **reference_video**: string - Reference video (3-10s) for style/camera reference or video editing
- **video_reference_type**: string (default: feature) - 'feature' for style/camera reference, 'base' for video editing
  Options: feature, base
- **keep_original_sound**: boolean (default: true) - Keep original sound from the reference video
- **generate_audio**: boolean (default: false) - Generate native audio. Mutually exclusive with reference video
- **mode**: string (default: pro) - 'standard' generates 720p, 'pro' generates 1080p, '4k' generates 4K (no reference_video)
  Options: standard, pro, 4k
- **aspect_ratio**: string (default: 16:9) - Aspect ratio. Required when not using a start frame or video editing
  Options: 16:9, 9:16, 1:1
- **duration**: integer (default: 5) - Video duration in seconds (3-15). Ignored for video editing (base)
- **multi_prompt**: string - JSON array of shot definitions for multi-shot mode, e.g. [{"prompt":"...","duration":3}] (max 6 shots)

## Examples

### Pro (1080p) with Audio
A multi-shot cinematic trailer at the pro tier with native audio enabled
```json
{
  "model": "kwaivgi/kling-v3-omni-video",
  "input": {
    "mode": "pro",
    "prompt": "A Wes Anderson style movie trailer with symmetrical framing and pastel colors, ending on the full cast standing on the hotel steps",
    "duration": 15,
    "aspect_ratio": "16:9",
    "generate_audio": true
  }
}
```

### Standard (720p), No Audio
The cheapest tier for quick iteration, audio generation left off
```json
{
  "model": "kwaivgi/kling-v3-omni-video",
  "input": {
    "mode": "standard",
    "prompt": "A paper airplane gliding through a sunlit office, weaving between desks",
    "duration": 5,
    "aspect_ratio": "16:9",
    "generate_audio": false
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- mode defaults to pro (1080p) and generate_audio defaults to false — omitting both bills the pro/no-audio rate
- 4k mode is priced the same whether or not generate_audio is set, but does not support reference_video
- Adding audio raises the price about 25-33% at the standard/pro tiers
- Use multi_prompt for coherent multi-shot sequences instead of chaining separate generations
- video_reference_type=base edits an existing clip; =feature only borrows its style/camera motion

## Documentation
https://replicate.com/kwaivgi/kling-v3-omni-video