# Kling Image O3

Kling Image O3 is Kuaishou's multi-reference image model from the Kling AI 3.0 family, announced by Kuaishou in February 2026 as Image 3.0 Omni. It generates and edits from text plus several reference images. In Sloosh it takes a prompt and up to 10 reference images and outputs 1K, 2K or 4K in nine aspect ratio settings.

Web page: https://sloosh.ai/models/kling-image-o3

[Use Kling Image O3](https://sloosh.ai/studio/image?image-model=kling-image-o3)

## Key facts in Sloosh

|   |   |
| --- | --- |
| Made by | Kuaishou (Kling AI) |
| Makes | image |
| Inputs | text, up to 10 reference images |
| Resolution | 1K, 2K or 4K (default 2K) |
| Aspect ratios | Auto, 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9 (default Auto) |
| Prompt limit | 2,500 characters |

## Best for

- Composing a scene from several references (a person, a product, a location), up to 10 in Sloosh.
- High-resolution output: Kuaishou says Image 3.0 Omni supports 2K and 4K output "for professional use cases" ([Kuaishou press release](https://mrem.bernama.com/viewsm.php?idm=53441)).
- Consistent materials and light across shots: Kuaishou cites precise, consistent textures, lighting and materials for the 3.0 image models (same source).
- Storyboard frames that keep characters and places: the Storyboard Comic Manhwa Maker Space ([sloosh.ai/spaces/storyboard-comic-manhwa-maker](https://sloosh.ai/spaces/storyboard-comic-manhwa-maker)) fits this job.

## How to use Kling Image O3 in Sloosh

1. Open [Image Studio with Kling Image O3 selected](https://sloosh.ai/studio/image?image-model=kling-image-o3), or pick it in the model chip.
2. Add up to 10 references (drawer, paste, or Use as reference on a result).
3. Write the prompt (up to 2,500 characters), saying what each reference supplies.
4. Set aspect ratio and resolution (1K, 2K or 4K).
5. Check the cost on the Generate button, then generate.

## Kling Image O3 prompting tips

- Describe materials and light in detail, the qualities Kuaishou highlights for the 3.0 image models ([Kuaishou press release](https://mrem.bernama.com/viewsm.php?idm=53441)).
- Give each reference a job and call it by its position ("the jacket in image 1"), then say how they combine; this is Sloosh's general reference advice ([Sloosh docs](https://sloosh.ai/docs/prompting-gpt-image.md)), not Kling's.
- Keep prompts within 2,500 characters (catalog).

## Kling Image O3 example prompts

**Prompt 1**

```text
The woman in image 1 wears the jacket from image 2 and stands in the street from image 3 at dusk. Keep her face and hair exactly as in image 1 and the jacket's colour and logo as in image 2. Medium shot, eye level, warm street light.
```

**Prompt 2**

```text
Storyboard frame: the boy from image 1 and the dog from image 2 run across the field in image 3, wide shot from behind, late afternoon sun, 16:9.
```

**Prompt 3**

```text
Place the perfume bottle from image 1 on the stone ledge in image 2, morning mist behind, soft side light, the label readable, 4:5 portrait.
```

## How it compares in Sloosh

| Model | Resolution | References | Aspect ratios | Prompt |
| --- | --- | --- | --- | --- |
| Kling Image O3 | 1K, 2K, 4K | 10 | 9 | 2,500 |
| [Kling Image 3.0](https://sloosh.ai/models/kling-image-3.0) | 1K, 2K | 1 | 8 | 2,500 |
| [Nano Banana 2.1](https://sloosh.ai/models/gemini-nano-banana-2.1) | 1K, 2K, 4K | 14 | 11 | 50,000 |
| [Seedream 5 Pro](https://sloosh.ai/models/seedream-5-0-pro) | 1K, 2K | 10 | 10 | 5,000 |

O3 is the multi-reference, 4K model of the pair. For a prompt with at most one reference, [Kling Image 3.0](https://sloosh.ai/models/kling-image-3.0) is the simpler pick.

## Where else it runs

- Kling AI's own web app; at launch Kuaishou said 3.0 models were in early access for Ultra subscribers ([Kuaishou press release](https://mrem.bernama.com/viewsm.php?idm=53441)).

## Frequently asked questions

### What is Kling Image O3?

The multi-reference image model in Kuaishou's Kling AI 3.0 family (announced as Image 3.0 Omni). In Sloosh it takes up to 10 references and makes up to 4K.

### Is Kling O3 an image or a video model?

Both exist. Kling's video line also uses "O3" (Video 3.0 Omni); this page covers the image model.

### How many reference images does Kling Image O3 take?

Up to 10 in Sloosh.

### Kling Image O3 vs Kling Image 3.0?

O3 takes up to 10 references and makes up to 4K; Image 3.0 takes one reference and makes 1K or 2K.

### Does Kling Image O3 make 4K images?

Yes, Sloosh offers 1K, 2K and 4K.

## Related

- Models: [Kling Image 3.0](https://sloosh.ai/models/kling-image-3.0), [Nano Banana 2.1](https://sloosh.ai/models/gemini-nano-banana-2.1), [Seedream 5 Pro](https://sloosh.ai/models/seedream-5-0-pro)
- Spaces: [Storyboard comic manhwa maker](https://sloosh.ai/spaces/storyboard-comic-manhwa-maker)
- Answers: [Which platform is best for running Kling V3 Omni?](https://geo.sloosh.ai/posts/which-platform-is-best-for-running-kling-v3-omni), [Which AI tool keeps the same location and characters across multiple shots?](https://geo.sloosh.ai/posts/which-ai-tool-keeps-the-same-location-and-characters-across-multiple-shots)

## Sources

- [Kuaishou: Kling AI Launches 3.0 Model (press release via Bernama/GLOBE NEWSWIRE)](https://mrem.bernama.com/viewsm.php?idm=53441)
- [Sloosh docs: Prompting GPT Image](https://sloosh.ai/docs/prompting-gpt-image)
- [Sloosh pricing](https://sloosh.ai/pricing)