Saltar al contenido

Model guide

Kling 3.0 API: choosing a variant

The Kling 3.0 page groups eleven model IDs. What each variant accepts, the field rules the API enforces, and which one fits your input.

Hermes AI team5 min read

This article is available in English only.

The Kling 3.0 page groups eleven model IDs. They differ in two ways: the task (text to video, image to video or motion control) and the tier (Standard, Pro, Turbo Standard, Turbo Pro, and one 4K variant). Each one has its own input schema, so a request that works for one variant can be rejected by another.

This guide compares the variants using only those schemas: which fields each one takes, their defaults, and the rules the API checks before it accepts a task. It does not rank output quality or speed. For what each variant costs, see the Kling 3.0 pricing tab.

The variants at a glance

VariantModel IDYou provide
Pro text to videokling-v3-0-pro-text-to-videoA prompt
Standard text to videokling-v3-0-std-text-to-videoA prompt
Turbo pro text to videokling-v3-turbo-pro-text-to-videoA prompt
Turbo standard text to videokling-v3-turbo-std-text-to-videoA prompt
Pro image to videokling-v3-0-pro-image-to-videoA first frame, optional last frame
Standard image to videokling-v3-0-std-image-to-videoA first frame, optional last frame
Turbo pro image to videokling-v3-turbo-pro-image-to-videoA first frame
Turbo standard image to videokling-v3-turbo-std-image-to-videoA first frame
4K image to videokling-v3-0-4k-image-to-videoA first frame, optional end frame
Pro motion controlkling-v3-0-pro-motion-controlA character image and a driving video
Standard motion controlkling-v3-0-std-motion-controlA character image and a driving video

The page opens on Pro text to video. Pick another variant on the page, or link to it directly, for example Turbo pro image to video.

Text to video

All four text to video variants take a prompt of up to 2,500 characters, a duration from 3 to 15 seconds and an aspectRatio of 1:1, 16:9 or 9:16 (default 16:9). The Kling 3.0 Pro and Standard variants accept more controls than the Turbo ones:

FieldPro and StandardTurbo Pro and Turbo Standard
promptRequiredRequired
negativePromptUp to 2,500 charactersNot accepted
duration"3" to "15", default "5""3" to "15", default "5"
aspectRatio1:1, 16:9, 9:161:1, 16:9, 9:16
cfgScale0 to 1, default 0.5Not accepted
soundRequired boolean, default trueNot accepted
multiShot, shotTypeMulti-shot switch and modeNot accepted
multiPromptUp to 6 shotsUp to 6 shots; when sent, the main prompt is ignored

Note that duration is a string here ("5", not 5). Sending a number is rejected as an unsupported value.

Image to video

The image to video variants replace aspectRatio with a first frame. The image is a public HTTPS URL or a media:<uuid> reference from POST /api/v1/media.

  • Pro and Standard take firstImageUrl (required) and an optional lastImageUrl, plus everything the text to video Pro and Standard variants accept. They also take elementList, up to three element references. The cfgScale default differs: 0.5 for Pro, 0.8 for Standard.
  • Turbo Pro and Turbo Standard take firstImageUrl (required), an optional prompt, duration and multiPrompt. The schema lists the image limits: .jpg, .jpeg or .png, at most 50 MB, at least 300 px per side, aspect ratio between 1:2.5 and 2.5:1.

4K image to video

The 4K variant has its own field names and rules, so do not copy a Pro request into it unchanged:

  • The first frame is imageUrl and the optional last frame is endImageUrl.
  • duration is an integer from 3 to 15 (5, not "5").
  • sound defaults to false.
  • Send either prompt or multiPrompt, never both. multiPrompt takes up to 100 shots and requires multiShot: true; with multiShot: true, a prompt is refused.
  • The schema limits the image to .jpg, .jpeg or .png, at most 10 MB, at least 300 px per side, aspect ratio between 1:2.5 and 2.5:1.
POST /api/v1/jobs
{"model": "kling-v3-0-4k-image-to-video","input": {"imageUrl": "https://example.com/first-frame.png","prompt": "The camera pushes in slowly while steam rises from the cup","duration": 5}}

Motion control

The two motion control variants transfer the motion of a driving video onto a character image:

  • imageUrl (required): the character. Width and height between 300 px and 65,536 px, aspect ratio between 1:2.5 and 2.5:1.
  • videoUrl (required): the driving video, 3 to 30 seconds.
  • characterOrientation (required, default video): with image the output supports up to 10 seconds, with video up to 30 seconds.
  • Optional prompt, negativePrompt, elementList (up to three) and keepOriginalSound (default true).

Multi-shot rules

The Pro and Standard text to video and image to video variants can split one task into several shots. The API checks these rules before it accepts the task:

  1. Set multiShot to true.
  2. With shotType: "customize" (the default), send multiPrompt with at least one shot. Each shot has a prompt of up to 512 characters and an integer duration from 1 to 15.
  3. The shot durations must add up to the task duration. Five seconds as 2 + 3 passes; 2 + 2 is refused.
  4. On image to video, multi-shot cannot be combined with lastImageUrl.

With shotType: "intelligence", no shot list is required.

POST /api/v1/jobs
{"model": "kling-v3-0-pro-text-to-video","input": {"prompt": "A lighthouse keeper climbs the stairs at dusk and lights the lamp","duration": "5","aspectRatio": "16:9","multiShot": true,"shotType": "customize","multiPrompt": [{ "prompt": "Wide shot of the lighthouse on a cliff at dusk", "duration": 2 },{ "prompt": "Close-up of the keeper's hand turning the lamp on", "duration": 3 }]}}

Which variant to start with

Start from what you have:

  • Only text. A text to video variant. Pro or Standard if you want a negative prompt, cfgScale, the sound switch or custom multi-shot; Turbo if a prompt, duration and aspect ratio are all you set.
  • A still image. An image to video variant. Pro or Standard if you also have a last frame or element references.
  • A 4K first frame. The 4K variant, with its own field names.
  • A video of the motion you want. A motion control variant.

Then compare the price of the shortlisted variants on the pricing tab, and run one in the playground before you wire it into code. The Kling 3.0 two-shot prompt is a ready request to start from, and the tasks reference covers submitting and polling.

All prompt examples

Two-shot sequence with custom shots

A cyclist in a yellow rain jacket rides through a quiet coastal town at dawn, wet cobblestones, soft blue light

VídeoKling 3.0

kling-v3-0-pro-text-to-video