AIREITER

Kling v3 Omni (O3) AI Video Generator and API

Try Kling v3 Omni (O3) via playground or API. Generate 3–15s clips from text, images, first/last frames, or a reference video at 720p, 1080p, or 4K.

Run with API
Model type
24H Status Monitor
Success:100%
Input
(Required)
0/5000
Ideas:
Aspect Ratio
16:9
9:16
1:1
Output

Kling v3 Omni

5 prices

Kling

SpecificationCreditsPrice (USD)
720P
8
per second
$0.08
per second
720P · audio
10
per second
$0.1
per second
1080P
10
per second
$0.1
per second
1080P · audio
12.5
per second
$0.125
per second
4K
44
per second
$0.44
per second
Charged per output second. Native audio raises 720p/1080p. A reference video uses the video-input rate. feature follows Video Length; base follows the uploaded clip length. 4K is the same with or without audio.

Kling v3 Omni (O3) video generation

Kling v3 Omni is Kuaishou's all-in-one video model. Search results also call it Kling O3. On AIReiter one model ID covers text, multi-image reference, first/last frames, and reference-video workflows.

Text, image, or video input

Start from a prompt. Optionally add up to 7 stills or one reference clip. First/last frame is a separate mode and cannot be mixed with those uploads.

3–15 seconds, up to 4K

Text and image jobs take any whole second from 3 to 15. Reference clips must be 3–10 seconds. Output can be 720p, 1080p, or 4K.

Optional native audio

Audio is off by default. Turning it on raises 720p and 1080p. 4K stays one rate. Audio cannot be combined with a reference video.

Playground and API

The fields on this page are the same parameters sent as kling_v3_omni on POST /api/openapi/submit.

Kling O3 results from AIReiter

Clips generated through the AIReiter OpenAPI on 28 August 2026. Duration and pixels are from the delivered files. Credits follow the live table on this page.

Text to video, 720p

Prompt: A golden retriever running on a sunny beach, cinematic, natural lighting.
Request: Text/Reference, 720p, 16:9, 3 seconds, audio off.
Delivered: H.264 MP4, 1280x720, 3.04 seconds.
Charge: selected seconds × the 720P row in the table.

Restyle a reference video

Input: a 10-second 720p clip. Video Refer Type = feature (restyle). Video Length = 3 seconds.
Delivered: H.264 MP4, 1280x720, 3.04 seconds.
Charge: selected seconds × the video-input rate for that resolution.

Edit the original clip

Input: the same 10-second 720p clip. Video Refer Type = base (edit original). Video Length is hidden.
Delivered: H.264 MP4, 1280x720, 10.04 seconds.
Charge: uploaded seconds × the video-input rate, not the duration dropdown.

What to make with Kling v3 Omni

Pick the input you already have. Do not mix first/last frames with reference media.

Storyboard a shot

Upload a first frame, optionally a last frame, and describe the motion between them.

Keep a subject consistent

In Text/Reference add up to 7 images as <<<image_1>>>. Use Elements and @name for up to 3 named subjects.

Restyle an existing clip

Upload one reference video and keep Video Refer Type on feature. New length follows Video Length.

Edit the original clip

Switch Video Refer Type to base. Output length matches the upload (3–10 seconds).

Text, first/last frame, or reference video

First/last frames cannot be mixed with reference images or a reference video in one request.

Text / Reference

Prompt required. Optional: up to 7 images and one MP4/MOV (3–10s). Default for restyle is feature.

First / last frame

First frame required, last frame optional. Do not send a reference video in this mode.

Restyle vs edit original

feature (restyle) bills Video Length. base (edit original) hides Video Length and bills the source duration. Those labels are the control names on this page.

Kling v3 Omni parameters

Playground and API share the same validated fields. Live prices stay in the table, not in this list.

Resolution

720p, 1080p, or 4K. 720p is the default. Higher rows cost more per second except 4K, which is one rate with or without audio.

Duration

3–15 seconds for text and image jobs. Hidden when editing the original clip; then the upload length (3–10s) is billed.

Aspect ratio

16:9, 9:16, or 1:1. Ignored when a first frame or reference video is provided.

Reference media

Up to 7 images. One reference video, MP4/MOV, 3–10s, max 200MB. Native audio cannot be combined with a reference video.

Kling v3 Omni pricing

Use the live per-second table on this page. Copy below only explains which row applies. Failed tasks are not charged.
01

Text and image jobs

Charge = selected seconds × the 720P, 720P · audio, 1080P, 1080P · audio, or 4K row.

02

Restyle a reference video

Video Refer Type = feature. Charge = selected seconds × the video-input rate for that resolution. Native audio is ignored.

03

Edit the original clip

Video Refer Type = base. Charge = uploaded seconds × the same video-input rate. The duration dropdown is hidden and is not billed.

04

Worked example

A 3-second silent 720p text job uses the 720P row three times. A 10-second original-clip edit uses the video-input row ten times, even if a duration of 3 was still in the request body.

How to generate Kling O3 video

Same steps in the playground and the API.

01

Pick one mode

Text/Reference or First/Last Frame. Switching tabs clears uploaded images and video.

02

Describe the motion

Who moves, then camera and lighting. Stills as <<<image_1>>>. Named subjects as @name.

03

Set length and resolution

3–15 seconds for text and image jobs. For original-clip edit, skip length; billing follows the upload.

04

Submit and poll

POST /api/openapi/submit returns estimated credits. Poll POST /api/openapi/query with the same out_task_id.

Kling v3 Omni API

Both modes share one asynchronous contract on AIReiter. The model ID is kling_v3_omni.

Model ID

Send kling_v3_omni. Set params.type to all_reference or first_last_frame.

Submit

POST /api/openapi/submit with model, params, and a unique out_task_id.

Query

POST /api/openapi/query with the same out_task_id. Completed responses include credits_used and the output video URL.

Reference video billing

Send video_url and refer_type. feature bills video_length. base bills the probed source duration and ignores video_length.

Related AI video models

Compare Kling v3 Omni with other video models on AIReiter.

Kling v3 Omni FAQ

Kling O3 length, pricing, first/last frame, reference-video billing, and API.

/ 01

What is Kling O3 / Kling v3 Omni?

Kling v3 Omni is Kuaishou's unified video model. Kling O3 is the same model. On AIReiter it generates 3–15 second clips at 720p, 1080p, or 4K from text, images, first/last frames, or one reference video.

/ 02

How long can Kling v3 Omni generate?

Any integer from 3 to 15 seconds for text and image jobs. A reference video must be 3–10 seconds. Editing the original clip follows the upload length.

/ 03

How much does Kling v3 Omni cost?

Open the pricing table on this page. Text and image jobs use the silent or audio row. A reference video uses the video-input rate. Failed tasks are not charged.

/ 04

How is a reference video billed?

Restyle (feature) multiplies selected seconds by the video-input rate. Edit original (base) multiplies uploaded seconds by that rate and ignores the duration field.

/ 05

What is the difference between restyle and edit original?

Restyle treats the upload as style or motion reference and outputs Video Length. Edit original rewrites the uploaded clip; output length matches the source.

/ 06

Can I mix first/last frames with a reference video?

No. First/last frames cannot be combined with reference images or a reference video in the same request.

/ 07

Where is the Kling v3 Omni API documentation?

API docs: https://docs.aireiter.com/en/api-reference/videos/kling-v3-omni/generation. Use kling_v3_omni on POST /api/openapi/submit, then poll POST /api/openapi/query.

/ 08

Which API model ID should I use?

Use kling_v3_omni.