Kling O3

Kling O3 API. Reference-driven AI video: image references, first/last frame, or a reference video to restyle or continue, 3-15 seconds.

POST/v1/media/queue

from $0.143/second

Overview

Kling O3 — reference-driven video generation. Steer a clip with reference images, pin its first and last frames, or pass a reference video to restyle or continue. Async operation — submit a job, poll for completion.

Property	Value
Model ID	`kwaivgi/kling-o3`
Alias	`kling-o3`
Provider	Kling
Inputs	text-to-video, reference images, first/last frame, reference video
Aspect Ratios	16:9, 9:16, 1:1
Duration	3–15 seconds (3–10 seconds in reference-video mode)
Audio Generation	Supported (off by default)
Billing	Per job — priced when queued, charged only on success
Pricing	from $0.143/second

Usage

All media models use the async job queue. Submit a job, then poll for the result.

Step 1: Submit Job

Text-to-Video

options.params passes this model's own parameters (see Model parameters).

const res = await fetch('https://api.yepapi.com/v1/media/queue', {
  method: 'POST',
  headers: {
    'x-api-key': 'YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'kwaivgi/kling-o3',
    prompt: 'A cinematic drone shot over a rugged coastline at golden hour',
    options: {
      aspectRatio: '16:9',
      duration: '5',
      generateAudio: true,
      params: {
        mode: 'pro',
      },
    },
  }),
});
const { data } = await res.json();
// data.jobId — use this to poll for results

curl -X POST https://api.yepapi.com/v1/media/queue \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "kwaivgi/kling-o3", "prompt": "A cinematic drone shot over a rugged coastline", "options": {"aspectRatio": "16:9", "duration": "5", "generateAudio": true, "params": {"mode": "pro"}}}'

Reference Images

Pass an image via imageData (without imageRole) to use it as a reference for the subject, style or setting.

const res = await fetch('https://api.yepapi.com/v1/media/queue', {
  method: 'POST',
  headers: {
    'x-api-key': 'YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'kwaivgi/kling-o3',
    prompt: 'The character from the reference image walks through a neon-lit night market',
    options: {
      duration: '5',
      generateAudio: true,
    },
    imageData: {
      mimeType: 'image/jpeg',
      base64: '...', // Base64-encoded image
    },
  }),
});

curl -X POST https://api.yepapi.com/v1/media/queue \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "kwaivgi/kling-o3", "prompt": "The character from the reference image walks through a neon-lit night market", "options": {"duration": "5", "generateAudio": true}, "imageData": {"mimeType": "image/jpeg", "base64": "..."}}'

First / Last Frame

Set options.imageRole to first_frame to switch to first/last-frame mode: imageData becomes the opening frame, and an optional closing frame goes in options.params.last_frame_url as a public image URL.

const res = await fetch('https://api.yepapi.com/v1/media/queue', {
  method: 'POST',
  headers: {
    'x-api-key': 'YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'kwaivgi/kling-o3',
    prompt: 'Smooth transition from the opening frame to the closing frame, slow push-in',
    options: {
      aspectRatio: '16:9',
      duration: '5',
      imageRole: 'first_frame',
      params: {
        last_frame_url: 'https://example.com/last-frame.jpg',
      },
    },
    imageData: {
      mimeType: 'image/jpeg',
      base64: '...', // Base64-encoded image
    },
  }),
});

curl -X POST https://api.yepapi.com/v1/media/queue \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "kwaivgi/kling-o3", "prompt": "Smooth transition from the opening frame to the closing frame, slow push-in", "options": {"aspectRatio": "16:9", "duration": "5", "imageRole": "first_frame", "params": {"last_frame_url": "https://example.com/last-frame.jpg"}}, "imageData": {"mimeType": "image/jpeg", "base64": "..."}}'

Reference Video

Pass a public MP4 or MOV URL in options.referenceVideoUrl to restyle, edit or continue an existing clip.

const res = await fetch('https://api.yepapi.com/v1/media/queue', {
  method: 'POST',
  headers: {
    'x-api-key': 'YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'kwaivgi/kling-o3',
    prompt: 'Restyle this clip as a hand-drawn watercolor animation',
    options: {
      aspectRatio: '16:9',
      duration: '5',
      referenceVideoUrl: 'https://example.com/reference.mp4', // public MP4 or MOV URL
    },
  }),
});

curl -X POST https://api.yepapi.com/v1/media/queue \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "kwaivgi/kling-o3", "prompt": "Restyle this clip as a hand-drawn watercolor animation", "options": {"aspectRatio": "16:9", "duration": "5", "referenceVideoUrl": "https://example.com/reference.mp4"}}'

Step 2: Poll for Result

const status = await fetch(`https://api.yepapi.com/v1/media/status/${data.jobId}`, {
  headers: { 'x-api-key': 'YOUR_API_KEY' },
});
const { data: job } = await status.json();
// job.status — "pending" | "processing" | "completed" | "failed"
// job.result.video — { mimeType, base64 } when completed

curl https://api.yepapi.com/v1/media/status/JOB_ID \
  -H "x-api-key: YOUR_API_KEY"

Request Body

Parameter	Type	Required	Description	Default
`model`	`string`	Yes	`kwaivgi/kling-o3` or alias `kling-o3`	—
`prompt`	`string`	Yes	Text description of the video to generate	—
`options.aspectRatio`	`string`	No	One of: 16:9, 9:16, 1:1	model default
`options.duration`	`string` or `number`	No	`3` to `15` (seconds); `3` to `10` (seconds) in reference-video mode	`5`
`options.generateAudio`	`boolean`	No	Generate an audio track for the video (not in reference-video mode)	`false`
`options.imageRole`	`string`	No	`first_frame` switches to first/last-frame mode (`imageData` = first frame); omit to use `imageData` as a reference image	—
`options.referenceVideoUrl`	`string`	No	Public MP4 or MOV URL of a reference video to restyle, edit or continue	—
`options.params`	`object`	No	Model-specific parameters under their own names: `elements`, `last_frame_url`, `mode`, `multi_prompt`, `multi_shots`, `shot_type`. Unknown names return 400	—
`imageData.mimeType`	`string`	No	MIME type of input image (image/png, image/jpeg, image/webp)	—
`imageData.base64`	`string`	No	Base64-encoded reference image, or the first frame with `imageRole`	—

Options are validated when you submit: an unsupported duration, resolution or value returns 400 naming the allowed values.

Model parameters

Pass these inside options.params.

Parameter	Type	Allowed	Description	Default
`elements`	`array`	—	Kling element IDs to keep consistent across the clip (only image input, text-to-video, reference-video mode)	—
`mode`	`string`	`std`, `pro`, `4k`	Quality mode	`pro` (first/last-frame mode, reference-video mode); `std` (image input, text-to-video)
`multi_prompt`	`array`	up to 6 shots	Multi-shot script: list of `{ prompt, duration }` shots (only first/last-frame mode, image input, text-to-video)	—
`multi_shots`	`boolean`	—	Generate a multi-shot sequence (only first/last-frame mode, image input, text-to-video)	`false`
`shot_type`	`string`	`customize`, `intelligent`	Multi-shot planning: your shots (`customize`) or the model's (`intelligent`) (only image input, text-to-video)	`customize`
`last_frame_url`	`string`	public image URL	Public URL of the last frame (first/last-frame mode)	—

Features

Text-to-video and reference-image video
First/last-frame mode (imageRole: 'first_frame' + options.params.last_frame_url)
Reference video via options.referenceVideoUrl
Optional native audio (off by default)
Multi-shot sequences via options.params.multi_shots / multi_prompt

Pricing

Tier	≈ Cost/sec	5-second video
Without audio	$0.143	$0.72
With audio	$0.19	$0.96

Prices shown are for a 5-second clip at the default settings; longer clips cost proportionally more. The mode parameter (std/pro/4k) changes the price. Each job is priced when it's queued — the exact quote for your duration, resolution and options — and charged only if it succeeds. Failed or moderated jobs are free.

Info

Video jobs can take several minutes to complete. When capacity is busy, jobs wait in the queue (status stays pending or processing) — no action needed. Results retained for 1 hour.

Under the Hood

Kling O3

On this page