ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›THE CRAFT›CHEAT SHEETS›KLING CHEAT SHEET
CHEAT SHEET · KUAISHOU · TEXT-TO-VIDEOINTERMEDIATE

Kling Cheat Sheet: Prompting 4.0, 3.0 and 3.0 Omni, Camera and Motion Control

Built on Kling’s own guides. Kling says 4.0 rolled out to all users on 15 Oct 2026, and Kling still lists 3.0 and 3.0 Omni, so this sheet covers all three.

READ17 min
WORDS3,476
SECTIONS8
SOURCES9
TYPECHEAT SHEET
CHECKED1 OCT 26

To prompt Kling, describe the subject, how it moves and the scene, then add camera, light and mood; on 3.0 Omni and 4.0, also tag each uploaded file with @ and say what it is for. Steer the camera with Kling’s camera controls or a reference video, and copy a performance with Motion Control. Kling 4.0 rolled out to all users on 15 Oct 2026, and Kling still lists 3.0 and 3.0 Omni. Everything here comes from Kling’s own guides, blog and API reference, plus one press release from its parent company, Kuaishou.

KLING 4.0 — STATUS ON 3 NOV 2026

Kling 4.0 has launched. Kling’s own blog says “Since Kling 4.0 rolled out to all users on October 15”, and its API documentation now carries the banner “Kling 4.0 API Is Now Officially Live and Ready for Integration!” Kling’s parent company, Kuaishou, announced on 19 October that Kling 4.0 “officially launched on October 15”. Some older Kling pages, dated 30 September, still say the full model is “scheduled to launch in October 2026”; the newer pages replace them. Kling still lists 3.0 and 3.0 Omni next to 4.0 in its API menu.1

Kling itself still marks three 4.0 items as coming soon, not available yet: 10-bit HDR output at 1080p and 4K, extending an uploaded video “up to 2 minutes”, and, in its API manual, Kling 4.0 Flash, the faster version, which Kling describes as up to 20 seconds at 720p with no start-and-end frames or keyframes. Kling’s app guide already describes Flash as a choice, so check which one your account shows.2

What Kling says 4.0 changes: clips of 3 to 30 seconds (3.0: 3 to 15), up to 10 keyframes (still images you supply to fix what chosen moments of the clip look like; 3.0 takes only a start and an end frame), up to 15 reference files in one go, prompts of up to 8,000 tokens, and 720p, 1080p or 4K output. These are Kling’s claims.3

TL;DR — THE SHORT VERSION
  • The plain answer: describe the subject, how it moves, and the scene, in that order; write anything you do not want as a plain sentence inside the prompt. On 3.0 Omni and 4.0, also tag each uploaded file with @ and say what it is for.
  • Kling 4.0 has launched. Kling says it rolled out to all users on 15 Oct 2026. Kling still marks 10-bit HDR and 2-minute video extension as coming soon, and still lists 3.0 and 3.0 Omni.
  • Camera: name shots and moves in the prompt (“camera language”), or use Kling’s camera controls: six basic moves (horizontal, vertical, zoom, pan, tilt, roll) and four Master Shots.
  • Motion Control: upload a reference action video and a character image, and the character copies the movement and expressions.
  • Kling's formula: Subject (with description) + Subject Movement + Scene (with description) + optional camera language, lighting and atmosphere.
  • A separate negative_prompt field survives only on the legacy API, for kling-v2-5-turbo, kling-v2-6 and kling-v3.
  • Even there, Kling says to use the prompt: “It is recommended to supplement negative prompt via negative sentences within positive prompts”
  • The upgraded API has no separate field: “The prompt can include positive and negative descriptions”
  • Kling 3.0 and 3.0 Omni clips run 3 to 15 seconds, with native audio. Kling says 4.0 goes to 30.
VENDOR-DOCUMENTED

Every rule on this sheet is quoted from Kling’s own documentation, guides or blog, linked where it is used. The copy-paste prompts are this site’s own examples.

The formula

Subject (description) + Subject Movement + Scene (description) + (Camera Language + Lighting + Atmosphere) — optional

Kling AI, Text-to-video prompt guide, read at source 16 Sep 2026.

Template:

[subject], [subject description], [how it moves], in [scene], [scene description]. [camera language]. [lighting]. [atmosphere].

Worked example:

A ceramicist in a flour-dusted apron centres clay on a wheel, drawing the form slowly upward, in a north-facing studio at dusk. Slow push-in. Window light only. Quiet, focused atmosphere. No text or watermark on screen.

Kling 4.0’s own structure is longer, and adds pacing and sound: “Subject + Action + Setting + Camera + Lighting + Pacing + Sound”. Kling also says 4.0 prompts can now run “up to 8,000 tokens”. Use the 3.0 formula on 3.0 and 3.0 Omni, and the longer one on 4.0.4

Breakdown — what each part does

01

Subject

The main focus — “It can be people, animals, plants, objects, and so on”. Describe it.

02

Subject Movement

Its own part of Kling's formula. Say how the subject moves.

03

Scene

Where it happens, with a description.

04

Camera, lighting, atmosphere

Optional in Kling's formula. Add them when the shot depends on them.

05

Physics

Kling's older text-to-video guide warned “it is challenging to generate complex physical movements, such as the bouncing of a ball or the trajectory of a high-altitude throw”. Its June 2026 post on 3.0 says “Cloth dynamics, hair movement, fluid behavior, and contact collisions are simulated in real time.” Test physics-heavy shots rather than assuming either.

3.0 Omni and 4.0: reference prompts

The plain point: on Kling 3.0 Omni and 4.0 you can upload pictures, video clips and saved characters (Kling calls a saved character an element), and the prompt says what each one is for. Kling treats them as part of the prompt: “the images, videos, elements, and text you upload are all treated as prompts by VIDEO 3.0 Omni”.5

HOW TO WRITE THE REFERENCES, FROM KLING’S GUIDES
Part of the prompt3.0 Omni4.0
Naming a file3.0 Omni: a saved character is written @name inside the sentence (Kling’s examples: @Boxer A, @Grace); an uploaded picture appears as @Image.4.0: files are numbered for you in upload order: @Image1, @Image2, @Video1.
Saying what it is for3.0 Omni: Kling’s examples put the name straight into the scene: “@Grace sits on the sofa eating cookies as @Alan walks in holding @Samoyed.”4.0: give every file a role first: “@Image1 defines the scene. Preserve its spatial layout, furnishings, and lighting.”
Several shots3.0 Omni: Kling’s examples write “Shot 1 (3s): … Shot 2 (2s): …”, or time ranges such as “[00:00 - 00:02] Medium shot:”. A clip runs up to 15 seconds.4.0: multi-shot is on by default, so the model splits shots itself; Kling’s template uses time segments, “0–X s: ___”. A clip runs up to 30 seconds.
Sound3.0 Omni: write who says what in quotes, or an “Audio:” line per shot. A voice saved on a character is bound to it and does not need to be written again.4.0: give sound its own part: “Dialogue/voiceover + ambient sounds + music + sound effects”. Only a voice profile can be uploaded, not music or sound effects.
How many files3.0 Omni: up to 7 images and elements, or 4 if you add one video of 3 to 10 seconds.4.0: up to 15 in all: 10 images, 5 videos (30 seconds in total), 7 elements and 3 voice profiles.
Changing a clip3.0 Omni: Kling says video editing works as it did in its earlier O1 model.4.0: write “Edit @Video1. Change ___ to ___. Keep everything else unchanged.” Writing only “Reference @Video1” may start a new video instead.

Kling AI, Kling VIDEO 3.0 Omni Model User Guide, 6 Feb 2026, read at source 3 Nov 2026: “@Grace sits on the sofa eating cookies as @Alan walks in holding @Samoyed.”; “[00:00 - 00:02] Medium shot:”; “Once created, the voice is bound to the subject and does not need to be specified again in the input field.”; “When a video is provided, a total of up to 4 images/elements can be uploaded. If no video is provided, up to 7 images/elements can be uploaded.”; “You can upload one video with a duration between 3s and 10s”; “The video editing, prompt transformation, and other features in 3.0 Omni function the same as in O1.” · Kling AI, Kling 4.0 User Guide, 26 Oct 2026, read at source 3 Nov 2026: “Assets are automatically numbered in upload order, such as @Image1, @Image2, and @Video1.”; “@Image1 defines the scene. Preserve its spatial layout, furnishings, and lighting.”; “Multi-shot is enabled by default, allowing the model to automatically split shots and organize pacing based on the content.”; “Dialogue/voiceover + ambient sounds + music + sound effects”; “Up to 15 reference assets in total across images, videos, voice profiles, and elements”. · Kling AI API documentation, Kling 4.0 Video User Manual, read in a browser 3 Nov 2026: “Edit @Video1. Change ___ to ___. Keep everything else unchanged.”; “Using only ‘Reference’ may route the request into reference-based generation instead of editing.”; “Audio input is supported only as voice profiles”.

Kling’s own advice for 4.0, in short:

  • Say what each file is for. Kling: “If you provide only asset IDs without specifying their roles, the model will infer their intended use, making the results less predictable.”
  • Use concrete words. Kling’s example swaps “Premium look” for “Side backlighting outlines the object’s edges, the background is blurred, and fine matte grain is visible on the surface.”
  • Start short. Kling says longer prompts are not always better: begin with one sentence naming the element, scene, action and emotion, then add only the details that change the result.
  • Keyframes. Kling’s format for a first frame is “Use @Image1 as the first frame, then describe what happens next.” The Keyframes entry and Omni Reference cannot be combined in one task.
  • Storyboard as a reference. Say you want finished frames, or Kling warns that grid lines, title bars or notes from the storyboard sheet may appear.

Kling AI, Kling 4.0 User Guide, 26 Oct 2026, read at source 3 Nov 2026: “If you provide only asset IDs without specifying their roles, the model will infer their intended use, making the results less predictable.”; “Premium look” and “Side backlighting outlines the object’s edges, the background is blurred, and fine matte grain is visible on the surface.”; “Use @Image1 as the first frame, then describe what happens next.”; “The capabilities of the two entry points cannot be combined within a single generation task.” · Kling AI, Kling 4.0 Video User Manual, read in a browser 3 Nov 2026: “Are longer prompts better?” “Not necessarily.”; “otherwise grid lines, title bars, or Chinese annotations from the storyboard sheet may appear in the generated video.”

USE WHEN: KLING 3.0 OMNI — SAVED CHARACTERS IN A SHORT MULTI-SHOT SCENE (SWAP @Baker AND @Customer FOR YOUR OWN ELEMENT NAMES)
Shot 1 (3s): Wide shot, @Baker stands behind the counter of a small bakery at dawn, flour hanging in the air. Shot 2 (3s): Close-up, @Baker lifts a tray of bread from the oven and says, “Still warm.” Shot 3 (2s): Mid-shot, @Customer walks in from the street and smiles. Soft window light. No subtitles.
USE WHEN: KLING 4.0 — TWO PICTURES AND A CAMERA REFERENCE CLIP, EACH WITH A ROLE
@Image1 defines the scene. Preserve its layout, furniture and lighting. @Image2 defines the woman. Preserve her face, hairstyle and clothing. Reference the camera path and pacing of @Video1. 0–4 s: she walks to the window and stops. 4–9 s: she looks back toward the camera as the camera pushes in slowly. Audio: soft rain on the glass, one low piano note at the end. No subtitles.

Camera control: language and controls

The plain point: Kling gives you two ways to direct the camera. Camera language is words in the prompt: Kling’s guide lists “ultra-wide angle shots, bokeh (background blur), close-ups, telephoto shots, low-angle shots, high-angle shots, aerial views, and depth of field”, and says camera language “should be differentiated from camera motion control”. Camera motion control is a setting: Kling’s camera guide says it “now supports six basic camera movements, including ‘horizontal, vertical, zoom, pan, tilt, roll’”, set with “absolute commands” and a displacement amount.6

KLING CAMERA MOTION CONTROL, FROM KLING’S CAMERA GUIDE
ControlNames Kling uses
Six basic movesHorizontal · Vertical · Zoom · Pan · Tilt · Roll
Four Master ShotsMove Left and Zoom In · Move Right and Zoom In · Move Forward and Zoom Up · Move Down and Zoom Out

Kling’s camera guide does not say which model versions offer these controls, so check the panel in the version you are using.

Per shot, on 3.0. Kling VIDEO 3.0’s multi-shot mode takes up to 6 shots, and in Custom mode you set each shot’s duration, framing and perspective, what happens, and its movement: “Add movement instructions such as a push-in, pan, orbit, handheld feel, or tracking movement”. Kling’s advice is that “Each shot should add a distinct visual or narrative beat”.7

On 4.0. Kling says a reference video can set “Motion, camera movement, pacing”, and that 4.0 can edit “camera positions, camera movements” in an existing clip.8 Kling’s 4.0 manual lists the camera words it understands (“push-in, pull-out, pan, lateral camera movement, tracking shots, and orbit shots”) and says that when a reference video controls the camera, you should “explicitly state which parts of the camera path and pacing to reference, then specify the replacement subject and scene.”2

USE WHEN: KLING — ONE CLEAR CAMERA MOVE IN THE PROMPT
A cyclist in a yellow rain jacket rides through a narrow old-town street after rain, wheels throwing up a fine spray. Low-angle tracking shot that keeps pace beside the bike. Wet cobblestones reflect warm shop lights. Calm, early-morning atmosphere. No text or watermark on screen.

Motion Control: copy a performance

The plain point: Motion Control makes your character perform the movement in a video you upload. Kling defines it as a tool that “enables precise control of a character’s movements and facial expressions based on a reference image”: you “Upload the reference action video and character image you want to imitate”, or pick a motion from Kling’s library.9

  • One character, fully visible. Kling asks for a single character with the “entire body and head clearly visible and not obstructed.”
  • A clean reference clip. Kling advises against cuts, camera movement and very fast motion in the reference video.
  • Which way the character faces. By default, “Character Orientation Matches Video”: movement, expressions, camera movement and orientation all follow the clip. “Character Orientation Matches Image” keeps the image’s orientation; Kling lists it for 2.6, where “Camera movements and other elements can be customised through prompts.”
  • Versions. Kling lists VIDEO 3.0 Motion Control, which it says improves facial consistency over 2.6, and VIDEO 2.6 Motion Control. On 3.0, the element library carries only the face, not “clothing, hairstyle, makeup, or props”, so describe those in the prompt or show them in the image.

Kling AI, Motion Control user guide, 5 Mar 2026, read at source 12 Oct 2026. Kling’s 4.0 announcement, read the same day, and its 4.0 User Guide, read at source 3 Nov 2026, describe no 4.0 Motion Control feature.

USE WHEN: MOTION CONTROL — THE PROMPT ADDS WHAT THE CLIP DOES NOT
The same woman from the character image, wearing a navy linen shirt and silver hoop earrings, hair tied back, dancing in a sunlit rooftop garden with potted lemon trees, soft late-afternoon light.
TAKEAWAY

Words for the kind of shot, controls or a reference clip for the movement. When a character must copy a real performance, Motion Control does it; the prompt then only fills in clothes, setting and light.

Does it take a negative prompt?

Only on the legacy API — and Kling recommends writing negatives inside the prompt anyway.LEGACY

Kling AI API, Text to Video (legacy), read in a browser 16 Sep 2026 and re-read 12 Oct 2026 (the reference is JavaScript-only): negative_prompt is listed for the model enum kling-v2-5-turbo, kling-v2-6, kling-v3, “Cannot exceed 2500 characters”, and “It is recommended to supplement negative prompt via negative sentences within positive prompts”. Upgraded Text to Video: “The prompt can include positive and negative descriptions”.

So write exclusions as sentences in the main prompt — "No text or watermark on screen." — whichever API you use.

Corrected 16 Sep 2026: an earlier version of this sheet said the field exists on Kling 1.6 and 2.1 Master and is gone on 3.0 and 2.6, citing a third-party schema page. Kling's own legacy API reference lists it for kling-v2-5-turbo, kling-v2-6 and kling-v3. It also recommended "around five terms" in the field and said Stable Diffusion habits transfer cleanly; neither had a source, and both were removed.

On 4.0, Kling’s own guide goes further: “Negative prompts can increase the likelihood of errors.” It says only two negatives are truly necessary, no subtitles and no background music, because the model may add them by default.3

TAKEAWAY

On Kling, exclusions belong in the prompt as plain sentences. The separate field is legacy, and Kling itself points you back to the prompt.

What not to do

Do not rely on the separate negative field. It exists only on the legacy API.✕
Do not skip subject movement — it is its own part of Kling's formula.✕
Do not go over the prompt limit. The upgraded API: “Maximum length: 2500 characters.”✕
Do not assume a complex physical action will land first time. Test it.✕
Do not feed Motion Control a reference clip with cuts, camera movement or a half-hidden body.✕
Do not upload references without saying what each one is for (3.0 Omni and 4.0). Kling: the model will infer their use, so results are less predictable.✕
Do not plan around Kling 4.0 features marked “coming soon” (10-bit HDR, 2-minute extension, and Flash in the API manual) until Kling says they are out.✕
KLING 3.0 IN ONE LINE

Kling's 3.0 guide (6 Feb 2026): “The new model generates up to 15 seconds of continuous video, with a flexible duration ranging from 3 to 15 seconds.” That limit is for 3.0 only; Kling says 4.0 goes to 30 seconds (see the box at the top). Kling’s launch press release for 3.0 is dated 5 February 2026 and says 3.0 was “now available for exclusive early access to Ultra subscribers”. An earlier version of this sheet gave a release date of 7 February 2026, and until 1 Oct 2026 it said no separate release date had been found.

Kling AI, press release on the 3.0 launch, on GlobeNewswire, 5 Feb 2026, read at source 29 Sep 2026.

Free or paid? What the free plan allows, who owns the output, and when you need to pay: free vs paid AI image, video and music tools.

Where this model sits among the others: which models take a negative prompt, and what each vendor publishes as its formula.

How it compares on price and limits is on the video model comparison.

The checklist

Before you send it:

1 — Subject described?□
2 — Subject movement stated?□
3 — Scene described?□
4 — Camera, lighting, atmosphere added where the shot needs them, or a camera control set?□
5 — Exclusions written as sentences in the prompt?□
6 — Inside the prompt length Kling documents?□
7 — On 3.0 Omni or 4.0: every @ reference tagged, with its role written?□
SOURCES

Kling AI, Text-to-video prompt guide · Kling AI, VIDEO 3.0 user guide · Kling AI, 3.0 photorealistic video guide · Kling AI API, Text to Video (legacy) · Kling AI API, Text to Video All read at source on 16 Sep 2026.

Added 3 Nov 2026 (read at source or in a browser that day): Kling AI, Kling 4.0 at Work · Kuaishou press release on Kling 4.0, carried by Bernama · Kling AI, Kling 4.0 User Guide · Kling AI API documentation, Kling 4.0 Video User Manual · Kling AI, VIDEO 3.0 Omni Model User Guide · Kling AI, Inside Kling 4.0 (still says “scheduled to launch in October 2026”; an older page).

Re-read 13 Oct 2026 for 4.0’s status: Kling AI, Inside Kling 4.0 · Kling AI, Kling 4.0 vs 3.0 · Kling AI, Introducing Kling 4.0 · Kling AI API, Text to Video (legacy), model menu.

Added 12 Oct 2026: Kling AI, Inside Kling 4.0 · Kling AI, Camera Control Guide · Kling AI, VIDEO 3.0 Multi-Shot · Kling AI, Motion Control user guide, all read 12 Oct 2026; the Kling 4.0 posts, the text-to-video guide and the legacy API reference re-read the same day.

Added 1 Oct 2026: Kling AI, Introducing Kling 4.0 · Kling AI, Kling 4.0 vs 3.0, all read 1 Oct 2026 · Kling AI 3.0 press release (GlobeNewswire), read 29 Sep 2026.

Model versions and vendor documentation both move. Re-read the source before relying on a specific number.

Every other sheet in this set: all model cheat sheets.

Related on this site: First Frame and Last Frame in AI Video · Sound in AI Video

SOURCES
  1. 1Kling AI, Kling 4.0 at Work: Bringing Creative Vision into Production, 27 Oct 2026, read at source 3 Nov 2026“Since Kling 4.0 rolled out to all users on October 15, creative teams across film, advertising, and animation have been putting these capabilities to work.” · Kuaishou Technology, press release on Kling 4.0, carried by Bernama, 19 Oct 2026, read at source 3 Nov 2026: “Kling 4.0, the new-generation video generation model of Kling AI, officially launched on October 15.” · Kling AI API documentation, Kling 4.0 Video User Manual, banner and model menu read in a browser 3 Nov 2026: “[New Release] Kling 4.0 API Is Now Officially Live and Ready for Integration!”; menu: Kling 4.0, Kling 3.0 Turbo, Kling 3.0 & 3.0 Omni, Kling O1, Kling 2.6 and Kling 2.5 Turbo. · Kling AI, Inside Kling 4.0: A Complete Guide to AI Video Creation, 30 Sep 2026, re-read at source 3 Nov 2026, not updated since: “The full Kling 4.0 model is scheduled to launch in October 2026.” ↑
  2. 2Kling AI API documentation, Kling 4.0 Video User Manual, read in a browser 3 Nov 2026“10-bit HDR output at 1080p and 4K [coming soon].”; “4.6 Video Extension(Coming Soon)” with “Upload a video and extend it multiple times, up to 2 minutes.”; “4.5 Kling 4.0 Flash (Coming Soon)” with “A single generation supports up to 20 seconds at a maximum resolution of 720p. First-and-last-frame and multi-keyframe inputs are not currently supported.” · Kling AI, Kling 4.0 User Guide, 26 Oct 2026, read at source 3 Nov 2026: “Kling 4.0 Flash is ideal for rapid creative iteration and short-form video creation.” ↑a“push-in, pull-out, pan, lateral camera movement, tracking shots, and orbit shots”; “explicitly state which parts of the camera path and pacing to reference, then specify the replacement subject and scene.” ↑b
  3. 3Kling AI, Kling 4.0 User Guide, 26 Oct 2026, read at source 3 Nov 2026“Kling 4.0 supports single-generation videos ranging from 3 to 30 seconds.”; “Up to 15 reference assets in total across images, videos, voice profiles, and elements”; “Up to 8,000 tokens”; “720p / 1080p / 4K”. ↑a“Negative prompts can increase the likelihood of errors.”; “The model may add these by default, so explicitly exclude them when they are not needed.” ↑b
  4. 4Kling AI, Inside Kling 4.0: A Complete Guide to AI Video Creation, 30 Sep 2026, read at source 12 Oct 2026 · Kling AI, Introducing Kling 4.0, 30 Sep 2026, read at source 12 Oct 2026“now up to 8,000 tokens”. ↑
  5. 5Kling AI, Kling VIDEO 3.0 Omni Model User Guide, 6 Feb 2026, read at source 3 Nov 2026“the images, videos, elements, and text you upload are all treated as prompts by VIDEO 3.0 Omni.” ↑
  6. 6Kling AI, Text-to-video prompt guide, 24 Nov 2025, re-read at source 12 Oct 2026 · Kling AI, Camera Control Guide, 24 Nov 2025, read at source 12 Oct 2026.↑
  7. 7Kling AI, Kling VIDEO 3.0 Multi-Shot, 28 Jul 2026, read at source 12 Oct 2026.↑
  8. 8Kling AI, Inside Kling 4.0: A Complete Guide to AI Video Creation, 30 Sep 2026, read at source 12 Oct 2026 · Kling AI, Introducing Kling 4.0, 30 Sep 2026, read at source 12 Oct 2026.↑
  9. 9Kling AI, Motion Control user guide, 5 Mar 2026, read at source 12 Oct 2026.↑

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
aipromptgeneer://$ about this siteAI PROMPT GENEERAI explained in plain English. Every claim links to its source, with the date it was checked.NEWSLETTERWhat changed in AI each week, in plain English.SUBSCRIBE FREE ↗AI PROMPT GENEER · MACHINE ASSISTANCE DISCLOSEDEVERY ARTICLE CARRIES ITS OWN CHECKED DATE