FLUX 3 Image
FLUX 3 Image combines text-to-image generation, editing with up to ten references and spatial control in one workspace. Draw rectangles on a chosen canvas, describe each element and choose how it should change. Grounding is available for factual subjects, and five resolution tiers range from 768sq to 4k.
Properties
- Text-to-image generation and editing with up to ten reference images.
- Bounding boxes for placement and Keep, Change, Move, Remove and Add operations.
- Five resolutions: 768sq, 1k, 1.5k, 2k and 4k. Optional web and image grounding.
How to use this model
- 1
Describe the scene, composition and exact text, or explain the changes to an existing image.
- 2
Add up to ten reference images for editing. Mention each reference using its displayed token.
- 3
Open the layout editor to draw and describe rectangles. Choose an explicit aspect ratio for the canvas.
- 4
For edits, choose whether to keep, change, move, remove or add each element. Boxes guide placement; they are not pixel masks.
- 5
Choose a resolution and review the quote before generating. Grounding can use web and image search for factual subjects.
Parameters
Prompt
promptDescribe the scene or a general edit. Optional when editing references with layout boxes.
Reference images
reference_imagesUp to ten images. Refer to them in their displayed order.
Aspect ratio
aspect_ratioChoose the output framing. Layouts require an explicit ratio.
Auto
21:9
2:1
16:9
3:2
7:5
4:3
5:4
1:1
4:5
3:4
5:7
2:3
9:16
1:2
9:21
Resolution
resolutionOutput resolution tier; exact dimensions vary with the aspect ratio.
768sq
1k
1.5k
2k
4k
Web grounding
groundingAllow web and image search to inform the generation.
Layout
layout_elementsPlace elements or target edits in the visual layout editor.