> 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/image-models/flux-3-image.md).

# Flux 3 Image

Black Forest Labs' image generation and editing model, built for layout control, precise multi-turn edits and native 4K output.

## Overview

Flux 3 Image is the image model of Black Forest Labs' Flux 3 family. It generates and edits in a single model, with explicit control over where each element sits in the frame.

Its edge is control: place elements by bounding box, compose from up to 10 references, and edit one area across several turns while the rest of the image stays pixel-identical.

### Strengths for marketers

* **Precise multi-turn edits**: Change one element, keep everything else. Pixels outside the edited area stay unchanged, so iterations don't drift.
* **Layout control**: Place products, headlines and props by bounding box instead of re-rolling until they land in the right spot.
* **Multi-reference composition**: Combine up to 10 reference images (product, model, setting, style) into one scene.
* **Native 4K**: Fine text and texture hold at print and large-screen sizes, with no separate upscaling step.

### Ideal use cases

* **Iterative retouching**: Swap a colorway, change a prop, fix a detail, without regenerating the whole visual.
* **Composed product & editorial visuals**: Hero shots, collages and layouts where placement carries the message.
* **Catalog variations**: Same scene, different SKU, with the rest of the frame locked.

### Weaknesses

* **Very small boxes can be ignored**: New elements requested in tiny areas sometimes don't appear. Give important elements room.
* **Boxes guide, they don't clip**: Placement is strong but not a hard mask. Check edges on tight layouts.

***

## How to use effectively

#### Key principles

* **Describe the scene, then the layout.** Start with a short caption of the whole image, then list the elements you want to place.
* **Place elements with bounding boxes.** Add the list of boxes at the end of your prompt. Each box is `[y_min, x_min, y_max, x_max]` on a 0–1000 grid, origin top-left, so the same layout works at any aspect ratio.
* **Edit one thing per turn.** Small, sequential edits keep the rest of the image intact. Feed each result into the next edit.
* **Say which reference does what.** Refer to references by order or by content: "the bag from the second image, on the marble table from the third image".

#### Example

```
Studio shot of a white sneaker on a pale concrete plinth, soft daylight from the left, minimal beige background.

[
  {"id": "sneaker", "description": "white leather sneaker, three-quarter view", "box": [380, 250, 720, 750]},
  {"id": "headline", "description": "bold black sans-serif text reading 'RUN LIGHT'", "box": [80, 150, 220, 850]}
]
```