Bria/

fibo_edit-1.5

Partner

$0.04

/ image

Instruction-based image editing powered by FIBO 1.5: edit one image with a text instruction or a mask, or compose up to four reference images into a single result.

Public

Paper

api

Input

Instruction

Text-based edit instruction (e.g. 'make the sky blue', 'add a cat'). This parameter serves as the text prompt.. (Default: empty)

ImageUpload Image File

Please upload an image file

You need to log in to use this model

Log In

Settings

Structured Instruction

A string containing the structured edit instruction in JSON format. Use this instead of `instruction` for precise, programmatic control.. (Default: empty)

Image 2Upload Image File

Please upload an image file

Image 3Upload Image File

Please upload an image file

Image 4Upload Image File

Please upload an image file

MaskUpload Image File

Please upload an image file

Aspect Ratio

The aspect ratio of the output image. Defaults to the aspect ratio of `image`. A single-reference edit was not trained to change it dramatically; multi-reference requests support all values.

Seed

Seed for deterministic generation. A random seed is used if omitted. (Default: empty, 0 ≤ seed ≤ 2147483647)

Negative Prompt

Concepts, styles or objects to exclude from the edited image.. (Default: empty)

Guidance Scale

How closely the edited image should follow the instruction. Default: 5 (Default: empty, 3 ≤ guidance_scale ≤ 5)

Steps Num

Number of diffusion steps. Default: 50 (Default: empty, 20 ≤ steps_num ≤ 50)

Output Type

The desired output image format.

Ip Signal

If true, returns a warning for potential IP content in the instruction.

Prompt Content Moderation

If true, returns 422 on instruction moderation failure. Default: true

Visual Input Content Moderation

If true, returns 422 on image or mask moderation failure. Default: true

Visual Output Content Moderation

If true, returns 422 on visual output moderation failure. Default: true

Output

Model Information

FIBO Edit 1.5 applies natural-language edits to an existing image, powered by Bria's FIBO 1.5 model. It keeps everything you did not ask to change — subject identity, lighting, colour tone and fine detail — and like every Bria model it is trained exclusively on licensed data, so the output is commercially safe to deploy.

Text and structured instructions

Describe the change in plain language and the model returns the edited image together with the structured_instruction JSON it derived from your words. Reuse that JSON with the same seed to reproduce an edit exactly, or adjust one field to refine it.

Masked edits

Supply a black-and-white mask the same size as the source to confine the edit to part of the frame: black areas are preserved, white areas are edited. Omit the mask and the instruction applies to the whole image.

Multi-reference composition

Pass up to four images to draw a subject, a style and a background from different sources into a single result. Multi-reference requests support every output aspect ratio; masks apply to single-image edits only.