Image models

ByteDance · Image generation & editing

Seedream5.0 Pro

Create an image.
Keep shaping it.

Start with a scene or a multilingual poster. Bring references together, make a precise edit, or work with individual transparent layers.

Models & API access Explore examples

Real Pro outputs. Every example includes its request JSON. Explore series generation with Seedream 5.0 Lite.

An orange ceramic vase, a green leaf and a lemon on a sunlit shelf.
From a single prompt request.json
Generate

Give a composition its first form.

Precise editing

Point to an object or select an area.

Multi-image

Bring two references into one scene.

Layers

Work with the objects individually.

Transparency

Edit a cutout and keep its background clear.

Typography

Create lettering across languages.

01 / Precise edits

Small changes.
New possibilities.

Try two ways to change the same scene. Drag across each image to compare the original with the result.

A point identifies an object. A box selects an area. Keep the rest of the scene in place.

Before: An orange ceramic vase, a green leaf and a lemon on a sunlit shelf.
After: The same still life with a cobalt-blue vase and the original lemon and leaf.
OriginalEdited

Change one object

Point to the vase and ask for cobalt blue. Keep the composition, lighting and surrounding objects.

Models & API access request.json

02 / Multi-image composition

Two references.
One composition.

Take the blue vase from the first image and use it to replace the lemon in the second. Keep the orange vase, green leaf and lighting in place.

Give each reference a clear role: the object to bring in, and the scene to preserve.

Models & API access request.json

Input · 2 images

The same still life with a cobalt-blue vase and the original lemon and leaf.
Image 1 · subject to copy
An orange ceramic vase, a green leaf and a lemon on a sunlit shelf.
Image 2 · scene to edit

Output

The orange vase and leaf remain; a smaller blue vase replaces the lemon.

03 / Layer decomposition

One scene.
Separate pieces.

Separate the background, vase, leaf and lemon. Each object comes back as its own transparent image, ready for your next edit.

Models & API access request.json
Base imageVaseLeafLemon

Toggle a layer to see what is underneath.

04 / Transparent editing

A new color.
The same transparency.

Take the extracted vase, change its color and keep its transparency. Bring the result into a product composition, a design or a new scene.

Start with a PNG that already has a transparent background. The edited result keeps that transparency, ready to place in your design.

Models & API access request.json

Input

The extracted terracotta vase with a transparent background.

Output

A cobalt-blue vase on a transparent background.

05 / Multilingual typography

Your words.
Part of the image.

Give the model the words, languages and layout. This exhibition poster brings English, French and Japanese into one composition. The lettering is generated with the image.

English
FORM & COLOR
French
FORME ET COULEUR
Japanese
かたちと色

Check spelling and layout before using generated text in your design.

Models & API access request.json
A Seedream Pro exhibition poster reading FORM & COLOR, FORME ET COULEUR and かたちと色.
Generated with Seedream 5.0 Pro · 1024 × 1536
English · French · Japanese

Bring these ideas into your app.

Download the reference images, seven request JSON templates and a Python helper. Prepare requests locally, then run them with your Mercatus API key. Generated results can vary.

Download example kit Models & pricing