Seven Versions · One Reference

All Kling AI Models, Documented Version by Version

Every model in the Kling API line on one page — Kling Video 3.0, 3.0 Omni, O1, 2.6, 2.5 Turbo, 2.1 and 2.0. Durations, resolutions, aspect ratios, native audio and motion control, with a page of detail behind each card.

7 Kling AI models3–15s duration range1080p documented ceiling

Where the numbers come from

Every duration, resolution and ratio below is drawn from publicly documented limits, not from marketing pages or our own testing. Providers change ceilings quietly, so confirm before you build against one.

Newest first, not best first

The cards run in release order. A newer release is not automatically the right pick — a fast 2.x version beats a flagship when you are still deciding what the shot even is.

What we are not

This is an independent reference. We do not resell access, we run no weights from this model family, and nothing on the page is a price list or an availability guarantee.

Pick a Kling Model

Newest first. Each card carries the headline specs; the model page behind it carries the prompt templates, limits and generation examples.

🔥 New

Kling Video 3.0

The smart-storyboard release — one prompt comes back as a directed multi-shot cut.

3–15s1080pnative audio
  • Smart storyboard plans the shots
  • 15s in a single generation
  • Legible native text in frame
🚀 Flagship

Kling Video 3.0 Omni

Consistency-first flagship: subjects, garments and voices survive a change of shot.

3–15s1080pmulti-reference
  • Omnipotent Reference 3.0
  • Video character subjects from a clip
  • Custom per-shot storyboards
Latest

Kling O1

The first unified multimodal model in the line — strongest at composing several references.

3–10s1080ppro mode
  • Video reference input
  • Multi-reference composition
  • Pro mode for detail retention
Popular

Kling 2.6

The version that made native audio ordinary — voice, effects and ambience in one pass.

5–10s1080pnative audio
  • Voice, SFX and ambience together
  • Motion and camera-path control
  • Audio-driven lip sync
Fast

Kling 2.5 Turbo

The throughput option — built for iterating on a shot list, not polishing one hero clip.

5–10s1080pturbo
  • Fastest turnaround in the line
  • High throughput for batches
  • Text and image to video
Stable

Kling 2.1

The dependable middle of the line — better prompt reading than 2.0, no audio surprises.

5–10s1080pstandard / pro
  • Enhanced semantic understanding
  • Standard and professional modes
  • Predictable motion behaviour
Legacy

Kling 2.0

The baseline the rest of the Kling API line is measured against — still fine for simple shots.

5–10s720p / 1080pstandard
  • Text and image to video
  • Lowest-complexity generations
  • Widest historical availability

Not sure which one?

Five common situations mapped to the version that answers each — story length, recurring faces, sound, speed and references.

decision guidespec table

Kling AI Models Compared at a Glance

The seven versions on the axes that actually change how you write a request: duration, resolution, ratios, audio, motion control and generation modes.

Specification comparison of all seven Kling AI video models
ModelDurationMax resolutionAspect ratiosNative audioMotion controlGeneration modes
Kling Video 3.03–15s1080p16:9 · 9:16 · 1:1 · 4:3 · 3:4Standard · Professional
Kling Video 3.0 Omni3–15s1080p16:9 · 9:16 · 1:1Standard · Professional
Kling O13–10s1080p16:9 · 9:16 · 1:1Standard · Pro
Kling 2.65–10s1080p16:9 · 9:16 · 1:1Standard · Professional
Kling 2.5 Turbo5–10s1080p16:9 · 9:16 · 1:1Turbo · Standard
Kling 2.15–10s1080p16:9 · 9:16 · 1:1Standard · Professional
Kling 2.05–10s720p / 1080p16:9 · 9:16 · 1:1Standard

Capability availability as publicly documented for each Kling version, reviewed 2026-08. Vendors change limits without notice — treat this as orientation, then confirm against your provider.

Decision Guide

Which Kling Model Should You Use?

Five situations that come up constantly, and the version of the model line that answers each one.

You need a story, not a shot

Several beats inside one clip — a setup, a turn and a payoff — without hand-specifying every camera position yourself.

Kling Video 3.0Smart storyboard schedules the shots and 15s gives them room.
The same face has to appear twice

A recurring presenter, a product held by the same hands, or a character who must be recognisable across a sequence of generations.

Kling Video 3.0 OmniReference 3.0 holds identity through angle and lighting changes.
The clip has to talk

Dialogue, a voiceover, sound effects or ambience that lines up with the picture, without a separate audio pass afterwards.

Kling 2.6Native audio and lip sync generated with the picture.
You are still exploring the idea

Twenty variations of a concept where turnaround matters more than the last ten percent of fidelity.

Kling 2.5 TurboFastest generations in the line, built for volume.
You are composing from references

A garment, a location plate and a product shot that all have to end up in one coherent frame.

Kling O1Unified multimodal model, strongest at multi-reference work.

Try the Workflow Before You Pick a Model

Every Kling API video job starts from a frame. Render one here for free, then decide which version of the model line you actually need.

Write one sentence

Subject, motion, lighting. The generator reads plain English and needs no account to run.

Get a real key frame

Our own image model renders the still — a genuine render, not a stock lookup, in your chosen ratio.

Walk the animation step

See the storyboard, render and encode stages a video generation job moves through, as a guided preview.

Seven Models, One Workflow

Whichever Kling version you land on, the Kling API workflow is the same: prompt, key frame, motion. Start it here and finish it in a full video render.

7 models documentedSpecs from public docsNo signup required
Start Creating Free

No signup · No credits · Runs in your browser

Not affiliated with Kling AI or Kuaishou. Kling is a trademark of its respective owner.