# Kling O1: one model for generating and editing video — references, background swaps, restyling and extension

> Kling O1 is a Kuaishou video model introduced on December 1, 2025 as the first unified multimodal video model. It generates from text, frames and up to 7 references and edits finished clips in words: swapping characters and backgrounds, changing weather and style. Clips run 3–10 seconds in 720p and 1080p. Kling 3.0 Omni replaced it in February 2026. Kling O1 is available in the AI2Face generator.

Developer: Kuaishou (Kling AI) · Released: 2025-12-01 · Updated 2026-10-28
URL: https://ai2face.com/models/kling-o1

## Key facts

| | |
|---|---|
| Developer | Kuaishou (Kling AI) |
| Release date | December 1, 2025; successor — Kling 3.0 Omni (February 2026) |
| Tasks | Generation from text, frames and references, editing, restyling, next shot, extension |
| Duration | 3–10 seconds |
| Resolution | 720p and 1080p |
| References | Up to 7 images or elements; up to 4 with a video |
| Video input | 1 clip of 3–10 s, up to 200 MB, no higher than 2K |
| Price in Kling | 6–8 credits per second without video input, 9–12 with video |
| In AI2Face | Yes, in the generator |

## What is Kling O1

Kling O1 is a video model from the Chinese company Kuaishou. Kling introduced it on December 1, 2025, kicking off its launch week, and called it "the world's first unified multimodal video model." Two days later came [Kling 2.6](/models/kling-2-6) with sound.

The idea behind O1 is one model instead of a toolbox. Text-to-video, frame-based generation, background swaps and restyling used to need separate modes. O1 is built on the MVL (Multi-modal Visual Language) concept: text, images, video and elements are described together in natural language, and the model works out the task itself.

**Kling IMAGE O1** came out alongside the video model — the same approach for images, from generation to precise editing.

In February 2026 O1 was succeeded by [Kling 3.0 Omni](/models/kling-3-0). An overview of the family is on the [Kling](/models/kling) page. On the Arena leaderboard as of September 21, 2026, Kling O1 Pro ranks 33rd of 48.

## What Kling O1 can do

**Generation from references.** Up to 7 images or elements: characters, objects, backgrounds. The model keeps them consistent throughout the clip.

**Elements.** An element is built from 4 images of one object from different sides — that way the model understands its shape better.

![A turnaround sheet of a toy robot: a red body, a round head, silver arms and a yellow antenna — front, side, back and three-quarter views](https://ai2face.com/_astro/element.CsMa3Bls_1hYk9R.jpg "Four angles of one object — that's how many images a Kling O1 element needs. Made with Nano Banana Pro")

**Editing a finished clip in words.** Kling O1 adds and removes objects and changes the character, background, color, weather and even the camera angle while keeping the source motion. It can also recolor individual elements, key out a green screen and add effects like fire or skin textures.

![The same frame of a street with a yellow tram and a woman under an umbrella: on the left a sunny summer day, on the right a snowy winter evening with streetlights](https://ai2face.com/_astro/weather-edit.DIsT6uVQ_Z1uHt6u.jpg "Changing the weather and time of day — a typical Kling O1 edit. In the winter version the model added a second tram — worth checking for discrepancies like this in edits too. Made with Nano Banana Pro")

**Restyling.** A clip is redrawn as anime, pixel art, claymation, watercolor, felt or impressionist painting.

![A ginger cat stretching on a blue sofa by a window: on the left a live-action frame, on the right the same frame in claymation style](https://ai2face.com/_astro/restyle.B7nRIkcM_PYROF.jpg "One frame in two styles: live action and claymation. Made with Nano Banana Pro")

**Next or previous shot.** From an uploaded clip, O1 generates what comes next or the shot before it, and can copy a camera move or an action from a reference.

**First and last frame.** Two images — the start and the end; the model builds the motion between them.

**Extension** of a finished clip.

## Kling O1 pricing

| Mode | Without video input | With video input |
|---|---|---|
| 1080p | 8 credits/s (5 s — 40, 10 s — 80) | 12 credits/s (5 s — 60, 10 s — 120) |
| 720p | 6 credits/s (5 s — 30, 10 s — 60) | 9 credits/s (5 s — 45, 10 s — 90) |

Prices are in Kling app credits. A video input makes each second 1.5 times more expensive.

## Kling O1 vs its successors

| | Kling O1 | [Kling 3.0 Omni](/models/kling-3-0) | [Kling 4.0](/models/kling-4) |
|---|---|---|---|
| Released | Dec 1, 2025 | February 2026 | Oct 15, 2026 |
| Length | 3–10 s | up to 15 s | 3–30 s |
| Resolution | 720p, 1080p | up to 4K | up to 4K (coming soon) |
| Audio | no | yes | stereo |
| References | 7 images or elements | 7 images or elements | up to 15: 10 images, 5 videos, 7 subjects |
| Video input | 1 clip, 3–10 s | 1 clip, 3–10 s | up to 5 clips, 30 s total |
| Storyboard | no | yes | 10 keyframes |

## What Kling O1 is good for

- **Swapping the background or character** in a finished clip without a reshoot.
- **Removing clutter** — passers-by, objects, lettering.
- **Changing the weather and time of day** in one scene.
- **Restyling** — live action into a cartoon or claymation.
- **Continuing a scene** — the next shot from a finished clip.
- **A clip with a character and a product** from 7 references.

## Kling O1 vs alternatives

| | Kling O1 | [Wan 3.0](/models/wan-3-0) | [MiniMax H3](/models/minimax-h3) | [HappyHorse](/models/happyhorse) |
|---|---|---|---|---|
| Developer | Kuaishou | Alibaba | MiniMax | Alibaba |
| Text-based clip editing | yes | yes | yes | yes (1.0 model) |
| Length | 3–10 s | up to 30 s | 4–15 s | 3–15 s |
| References | 7 images or elements | 10 photos, 5 videos, 5 audio | 9 photos, 3 videos, 3 audio | up to 9 photos |
| Audio | no | yes | stereo | yes |
| Arena rank (Sep 21, 2026) | 33 | 6 | 8 | 10 (1.0) |

O1 was the first to bring generation and editing into one model, but today its neighbors do the same with sound and longer clips. It's worth using for short edits without sound; for new clips, 3.0 Omni or Wan 3.0 are more convenient.

## How to write prompts for Kling O1

Refer to files with `@`: `@Image`, `@Image1`, `@Image2`, `@Video`, `@Element`.

| Task | Template |
|---|---|
| Add | Add [what] from @Image to @Video |
| Remove | Remove [what] from @Video |
| Replace a character | Change [who] in @Video to @Element |
| Change the background | Change the background in @Video with @Image |
| Change the style | Change @Video to [style] style |
| Next shot | Based on @Video, generate the next shot: [description] |
| Frames | Take @Image1 as the start frame, take @Image2 as the end frame, [what happens] |

For reference-based generation, Kling suggests this order: element descriptions → how they interact → environment → lighting and style. For text-to-video: subject → motion → scene → camera, lighting and atmosphere.

```
Change @Video to claymation style: visible fingerprints, soft shadows,
slightly choppy movement like stop-motion. The cat, the sofa and the plant
on the windowsill stay in their places.
```

## How to use Kling O1 in AI2Face

Kling O1 is in the AI2Face generator alongside [Kling 3.0](/models/kling-3-0) and [Kling 4.0](/models/kling-4).

1. Sign up — new accounts get starter credits.
2. Open the generator, switch to video and pick Kling O1.
3. Upload a clip to edit or reference images and write a prompt that refers to them.
4. The credit cost is shown before you start. Credits don't expire.

Angles for an element are easy to make in the same place — with [Nano Banana Pro](/models/nano-banana-pro), like the robot sheet above.

## Limitations

- **No sound.** Native audio arrived with Kling 2.6 and 3.0 Omni.
- **3–10 seconds.** Longer only via extension or in newer models.
- **One input video,** 3–10 seconds, up to 200 MB and no higher than 2K.
- **Fewer references with a video** — up to 4 images or elements instead of 7.
- **You can't set only the last frame** — a start frame is required too.
- **Images** — JPG or PNG, at least 300 px, up to 10 MB.
- **Previous-generation model** — 33rd on Arena.

## Bottom line

Kling O1 is where editing video in words started at Kling: background and character swaps, weather and style changes, the next shot — all in one model. For short edits without sound it's still handy and cheaper than its successors. For new clips with sound and storyboards, use [Kling 3.0](/models/kling-3-0); for 30 seconds and 10 keyframes, [Kling 4.0](/models/kling-4); for long clips from documents, [Wan 3.0](/models/wan-3-0). All of them are in the AI2Face generator.

## FAQ

### What is Kling O1?

A Kuaishou video model that both generates and edits clips in a single model. Kling called it the first unified multimodal video model: text, images, video and elements are understood together.

### How do you edit video in Kling O1?

Upload a clip and say what to change, referring to it as @Video: "Remove the passers-by from @Video," "Change the background in @Video to @Image." The model changes what you specify and keeps the source motion.

### How is Kling O1 different from Kling 3.0 Omni?

3.0 Omni is O1's successor: clips up to 15 seconds instead of 10, native audio, shot-by-shot storyboards, characters with voices and editing up to 4K. The reference syntax stayed similar.

### How much does Kling O1 cost?

In the Kling app it's 8 credits per second at 1080p and 6 at 720p. With a video input it's 12 and 9. A 10-second 1080p clip without video input costs 80 credits.

## Sources

- https://ir.kuaishou.com/news-releases/news-release-details/kling-o1-launches-worlds-first-unified-multimodal-video-model-0
- https://kling.ai/quickstart/klingai-video-o1-user-guide
- https://kling.ai/quickstart/klingai-image-o1-user-guide
- https://kling.ai/quickstart/klingai-video-3-omni-model-user-guide
- https://arena.ai/leaderboard/text-to-video