> For the complete documentation index, see [llms.txt](https://docs.pletor.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.pletor.ai/models/video-models/kling-4.0.md).

# Kling 4.0

Kling's latest video model, for generation (Kling 4.0) and video editing (Kling 4.0 Edit).

## Overview

**Evolution from** [**Kling 3.0**](https://docs.pletor.ai/models/video-models/kling-3.0.md)**:** longer single-pass generation (3–15s → 3–30s), stereo audio with better lip-sync, dialogue and on-screen text in more languages, a much larger reference budget (up to 10 images and 5 videos), and 21:9 output.

### Strengths

* **Up to 30-second one-takes**: long takes and continuous scenes in a single generation, set to the second from 3 to 30 seconds.
* **Omni references**: combine up to 10 reference images and 5 reference videos in one run, cited as `@Image1`, `@Video1`. A reference video can drive performance, action, camera movement and pacing, or keep its composition while you swap the product or character.
* **Stereo audio and lip-sync**: dialogue, ambience and effects in stereo, with tighter lip-sync than Kling 3.0. Dialogue now also in Portuguese, German, French and Hindi.
* **Steadier motion and text**: better stability on fast action and big camera moves; logos and captions stay clearer.
* **First / last frame**: animate from a start image, optionally landing on an end image.
* **Video editing** (Kling 4.0 Edit): rewrite a 3–30 second clip from a text instruction (expressions, movement, characters, backgrounds, camera, style, weather, materials), keeping its length and framing.

### Ideal use cases

* **UGC-style content**: longer dialogue-driven takes with natural sound.
* **Product videos**: a product from reference images, consistent across a 30-second scene.
* **Character videos**: recurring characters, with performance borrowed from a reference video.
* **Post-production**: swap a background, outfit or product in existing footage with Kling 4.0 Edit.

### Weaknesses

* **1080p maximum**: for native 4K, use Kling 3.0 4K or upscale.
* **No middle keyframes, Elements or voice references in Pletor yet**: for Elements-based identity locking, use Kling 3.0 Elements.
* **Edits keep the source's length and aspect ratio.**

***

## How to use effectively

#### Model versions

* **Kling 4.0**: Reference mode (default): text-to-video, or up to 10 images and 5 videos (each 3–30s, 30s combined); **Keep audio** reuses the first reference video's soundtrack. First / last frame mode: a start frame (at least 300px a side, ratio 0.4–2.5) with an optional end frame; with no image, it runs text-to-video.
* **Kling 4.0 Edit**: source clip 3–30 seconds, at least 700px a side, 24–60 fps; up to 10 reference images; Keep audio on by default.

**Settings**: duration 3–30 seconds (edits match the source); 720p or 1080p; 16:9, 9:16, 1:1, 21:9 or Auto.

**Prompting**: name every reference by its tag, in connection order: `@Image1 is the bottle; keep its label exactly. Use the camera move from @Video1.` In edits the source is `@Video1`: describe only the change and say what must stay. Script dialogue with speaker, language and tone before each line.
