
BLACKFORESTLABS FLUX 3 Edit Image API by Black Forest Labs
Make precise local edits without changing the rest of the image, or combine up to 10 reference images into one well-composed result, with FLUX 3.
BLACKFORESTLABS FLUX 3 Edit Image is developed by Black Forest Labs. Atlas Cloud (operated by Atlas Cloud AI LLC) provides access to it and does not own it. All trademarks belong to their respective owners.
FLUX 3 Edit Image
FLUX 3 Image Edit is Black Forest Labs' newest image editing model. Make precise, local edits — add, replace, remove or restyle a single element — without changing the rest of the image, or combine up to 10 reference images into one well-composed result.
Why Choose This?
- Precise local editing — change one object, color or region; identity, pose, lighting, background and framing all stay as they were.
- Multi-reference composition — combine up to 10 references while keeping characters, products and objects recognizable.
- Reference-based style transfer — apply the look of one image to the content and composition of another.
- Text editing — change or add text inside an image.
- One flat price per image, whatever the number of references.
Parameters
| Parameter | Required | Description |
|---|---|---|
model | Yes | black-forest-labs/flux-3/edit |
prompt | Yes | Describes the edit. |
reference_images | Yes | 1 to 10 image URLs, in image 1 through image 10 order. Each image must be at least 256 px per side; images larger than 4 MP are downscaled, keeping aspect ratio. |
aspect_ratio | No | 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. Default auto, which matches image 1. |
resolution | No | 512sq, 768sq, 1k, 2k, 4k. Default 1k. Sets the price. |
enable_prompt_expansion | No | Clarifies the prompt and makes each reference's role explicit. Default false. |
output_format | No | jpeg or png. Default jpeg. |
safety_tolerance | No | 0 (strictest) to 4 (most permissive). Default 2. |
How to Use
curl -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \ -H "Authorization: Bearer $AIP_API_KEY" \ -H "Content-Type: application/json" \ --data-raw '{ "model": "black-forest-labs/flux-3/edit", "prompt": "Replace the background in image 1 with the fabric pattern from image 2", "reference_images": [ "https://static.atlascloud.ai/media/images/example-subject.png", "https://static.atlascloud.ai/media/images/example-pattern.png" ], "resolution": "1k", "output_format": "png" }'
Generation is asynchronous — poll /api/v1/model/prediction/{request_id} until status is completed, then read the image URL from outputs.
Prompting Tips
- Name the one change. Short and direct works best: "Turn the white heart into a red heart".
- Be specific about which element: "change the color of only one bird in the middle to #e01075", not "change a bird's color".
- Refer to references by number. Image 1 is the first entry in
reference_images: "Replace the sunglasses in image 1 with the round pink-tinted sunglasses from image 3". - For style transfer, say which image supplies what: "Turn image 1 in the style of image 2".
- Make one edit per request; chain several small edits instead of one large rewrite.
Notes
- Order matters — the first entry in
reference_imagesis image 1, andaspect_ratio: "auto"follows it. 4kcan take several minutes; the lower tiers return much faster.- Generated images are re-hosted on AtlasCloud storage, so the returned URL is ours, not the upstream provider's.
Related Models
- FLUX 3 Text-to-Image — Generate an image from a text prompt alone.
- FLUX 3 Edit Video — The same instruction-driven editing, for video.



















