Kling 3.0

1 models available

Provider: KLING

Updated: Recently

Available Models

Kling 3.0 T2V

kling-3.0-t2v

Experience the power of Kling 3.0 T2V for your AI applications.

Kling · Kling 3.0 Series

Kling 3.0 API: Multi-Shot AI Video, Turbo Speed, Omni Control

Eight models, three variants, one endpoint — from text-to-video to 4K video editing.

Kling 3.0 turns prompts, frames, reference images and source videos into 3-15 second clips. Script up to six shots in a single request, lock characters with @element references, add native audio, and step up to Omni for reference-to-video, motion reference and video editing at up to 4K.

Get API Access Async Task Docs View Pricing

Variants

3 tiers · 8 models

Standard, Turbo and Omni on the same /v1/videos/generations endpoint

Duration

3-15 seconds

Default 5s; multi-shot requests can derive duration from the shot list

Multi-Shot

Up to 6 shots

Per-shot prompts and durations in one request (Standard up to 5)

Resolution

Up to 4K

720p / 1080p on every variant, 4K on Omni

Standard, Turbo or Omni — Pick Your Variant

All three variants share the same request shape on APIPod. Switching between them is a one-line change to the model field.

Kling 3.0

Multi-shot storytelling with reusable elements

kling-3.0-t2v · kling-3.0-i2v

Text-to-video and first / last frame image-to-video (1-2 images)
Multi-shot: up to 5 shots, 1-12 seconds each, with per-shot prompts
Up to 3 @elements, each built from 2-4 images, referenced by name in the prompt
Optional native audio — off by default for single shots, on by default for multi-shot

Kling 3.0 Turbo

Fast generation with native audio built in

kling-3.0-turbo-t2v · kling-3.0-turbo-i2v

Text-to-video and single first-frame image-to-video
Every generation includes native audio
Shot lists of up to 6 shots via multi_prompt
Optional watermarked copy via the watermark flag

Kling 3.0 Omni

Reference, motion and editing control up to 4K

kling-3.0-omni-t2v · -i2v · -r2v · -edit

Reference-to-video with 1-7 reference images, addressable as @image_1…@image_N
Motion reference: one feature video (plus up to 4 images) drives the movement
Video editing on a 3-15.5 second MP4 / MOV source, with an option to keep its original sound
720p, 1080p and 4K output, plus automatic or scripted multi-shot

Kling 3.0 Variant Comparison

Same series, same unified API — different trade-offs between control, speed and output options.

Specification	Kling 3.0	3.0 Turbo	3.0 Omni
Task modes	T2V, I2V	T2V, I2V	T2V, I2V, R2V, Edit
Resolution	720p (default) / 1080p	720p (default) / 1080p	720p (default) / 1080p / 4K
Duration	3-15 seconds	3-15 seconds	3-15 seconds (3-10 with motion reference; edit follows source)
Image-to-video frames	First + optional last frame	First frame only	First + optional last frame
Multi-shot	Up to 5 shots, 1-12s each	Up to 6 shots	Up to 6 shots, or automatic storyboarding
References	Up to 3 @elements (2-4 images each)	—	1-7 reference images, or 1 motion video + up to 4 images
Native audio	Optional	Always on	Optional (not with video input)
Watermarked copy	—	Optional	Optional

All variants are served through POST /v1/videos/generations with task polling. Exclusions belong in the prompt — negative_prompt is not supported.

Capabilities

What Kling 3.0 Can Do

From scripted multi-shot sequences to editing existing footage — every capability maps to a documented request field.

Multi-Shot Storytelling

Pass multi_prompt as a list of {prompt, duration} shots to direct a sequence in one request. Kling 3.0 takes up to 5 shots; Turbo and Omni take up to 6 that add up to the clip duration.

Standard · Turbo · Omni

Element References (@name)

Define up to 3 kling_elements — a name, a description and 2-4 images each — then mention them as @name in the prompt to keep characters and props consistent.

kling-3.0-t2v · kling-3.0-i2v

Omni Reference Images

Send 1-7 reference images to kling-3.0-omni-r2v. Images are numbered automatically, so the prompt can point at @image_1, @image_2 and so on.

kling-3.0-omni-r2v

Motion Reference

Attach one feature video in video_urls (plus up to 4 images) to transfer its movement. With a motion reference, duration is 3-10 seconds and only the output length is billed.

kling-3.0-omni-r2v

Video Editing

Edit a 3-15.5 second MP4 / MOV source video with a prompt and up to 4 reference images. Set keep_original_sound to retain the source audio; aspect ratio follows the source.

kling-3.0-omni-edit

4K Output

Every variant renders 720p or 1080p. Omni adds a 4K resolution option for final deliverables.

Omni

Native Audio

Turn on generate_audio to render synchronized sound with the video. Turbo always includes audio; on Omni, audio is not available when a video input is used.

Standard · Turbo · Omni

Eight Models, One Endpoint

Every model is an async task on POST /v1/videos/generations — submit, poll, retrieve the video URL.

kling-3.0-t2v

Kling 3.0 · Text to Video

Prompt-driven generation with multi-shot, @element references and optional native audio. 16:9, 9:16 or 1:1.

Input: prompt (or multi_prompt)

kling-3.0-i2v

Kling 3.0 · Image to Video

Animate a first frame, or interpolate between a first and last frame. Aspect ratio follows the first frame.

Input: 1-2 frame images + prompt

kling-3.0-turbo-t2v

Turbo · Text to Video

Fast text-to-video that always renders native audio, with optional shot lists of up to 6 shots.

Input: prompt (or multi_prompt)

kling-3.0-turbo-i2v

Turbo · Image to Video

Animate a single first frame with native audio; the prompt is optional.

Input: 1 first-frame image + optional prompt

kling-3.0-omni-t2v

Omni · Text to Video

Text-to-video up to 4K with automatic or scripted multi-shot and optional native audio.

Input: prompt (or multi_prompt)

kling-3.0-omni-i2v

Omni · Image to Video

First frame or first + last frame image-to-video at up to 4K; the prompt is optional.

Input: 1-2 frame images + optional prompt

kling-3.0-omni-r2v

Omni · Reference to Video

Keep subjects consistent with 1-7 reference images, or drive movement with one motion reference video plus up to 4 images.

Input: prompt + 1-7 images, or 1 video + up to 4 images

kling-3.0-omni-edit

Omni · Video Editing

Edit an existing 3-15.5 second clip with a prompt and up to 4 reference images, optionally keeping the original sound.

Input: prompt + 1 source video + up to 4 images

Quickstart

Script a Multi-Shot Clip in One Request

One async endpoint for all eight models. Submit the task, poll the status, pick up the video URL — no SDK required.

Shot durations for kling-3.0 must add up to no more than duration; for Turbo and Omni they must add up to exactly duration. Prompts are limited to 2,500 characters.

import os, requests

payload = {
  "model": "kling-3.0-t2v",   # or -turbo-t2v / -omni-t2v ...
  "resolution": "1080p",     # 720p / 1080p (Omni: 4k)
  "duration": 10,            # 3-15
  "aspect_ratio": "16:9",    # 16:9 / 9:16 / 1:1
  "generate_audio": True,
  "multi_prompt": [
    {"prompt": "Wide shot: a lighthouse on a cliff at dusk, waves crashing", "duration": 4},
    {"prompt": "Close-up: the keeper lights the lamp, warm glow on his face", "duration": 6}
  ]
}

resp = requests.post(
  "https://api.apipod.ai/v1/videos/generations",
  json=payload,
  headers={"Authorization": f"Bearer {os.environ['APIPOD_API_KEY']}"},
)
task_id = resp.json()["data"]["task_id"]

# poll GET /v1/videos/status/{task_id} until status == "completed"

Frequently Asked Questions

What is the Kling 3.0 series on APIPod?

Kling 3.0 is a family of eight video models in three variants: Kling 3.0 (kling-3.0-t2v, kling-3.0-i2v), Kling 3.0 Turbo (kling-3.0-turbo-t2v, kling-3.0-turbo-i2v) and Kling 3.0 Omni (kling-3.0-omni-t2v, -i2v, -r2v and -edit). All of them run through APIPod's unified async endpoint POST /v1/videos/generations.

Which variant should I choose?

Choose Kling 3.0 when you want multi-shot storytelling with reusable @element references and optional audio. Choose Turbo when every clip should include native audio. Choose Omni when you need reference images, a motion reference video, video editing or 4K output.

How does multi-shot generation work?

Send multi_prompt as a list of {prompt, duration} shots. Kling 3.0 accepts up to 5 shots of 1-12 seconds each (up to 500 characters per shot) whose total does not exceed duration. Turbo and Omni accept up to 6 shots (up to 512 characters each) that add up to exactly duration. On Omni you can also set multi_shot without multi_prompt to let the model storyboard automatically. If duration is omitted, it is taken from the sum of the shot durations.

What are element references?

kling-3.0-t2v and kling-3.0-i2v accept up to 3 kling_elements, each with a name, a description and 2-4 images. Mention an element as @name in the prompt to keep that character or object consistent. Each @name reference counts as 37 characters toward the prompt length limit.

What can Omni do with reference images and videos?

kling-3.0-omni-r2v takes 1-7 reference images, or one motion reference video plus up to 4 images (duration 3-10 seconds). kling-3.0-omni-edit takes exactly one 3-15.5 second MP4 / MOV source video plus up to 4 reference images, and can keep the original sound. Inputs are numbered automatically, so prompts can reference @image_1 or @video_1.

Does Kling 3.0 generate audio?

Yes. Set generate_audio to true on Kling 3.0 and Omni. On Kling 3.0 it defaults to off for single-shot and on for multi-shot requests; on Omni it defaults to off and is not available when a video input is used. Turbo always generates audio.

How is Kling 3.0 billed?

Billing is per second of output, with rates that depend on resolution and — for Kling 3.0 and Omni — on audio or video input. kling-3.0-omni-edit is billed on the duration of the source video, and a motion reference video is not added to the billed duration. See the pricing page or the model page for current rates.

Which API do I call?

POST https://api.apipod.ai/v1/videos/generations with the model set to one of the eight kling-3.0 model IDs. The endpoint returns a task_id; poll GET /v1/videos/status/{task_id} until the status is completed, or pass callback_url to receive the final result. Then read the video URL from the result.

Start Building with Kling 3.0

Script multi-shot scenes, keep characters consistent and edit footage up to 4K — on the same unified API as the rest of your stack.

Get Your API Key View Pricing