Generate and edit in one place
Write a prompt alone, or add up to 10 reference images to edit, combine, or restyle them.
Try Seedream 5.5 online or via API. Generate and edit images from text or up to 10 references, export transparent PNGs, and split images into layers.
| Specification | Credits | Price (USD) |
|---|---|---|
| Model & ModalitySeedream 5.51K / 1.5K | 1.8per image | $0.018per image |
| Model & ModalitySeedream 5.52K | 3.15per image | $0.0315per image |
| Model & ModalitySeedream 5.5 Layerize1K / 1.5K (per output image) | 2.52per image | $0.0252per image |
| Model & ModalitySeedream 5.5 Layerize2K / Auto (per output image) | 3.87per image | $0.0387per image |
Prices follow the model configuration. Reference images are not billed unless the model says so.
One model for generation and editing, plus a Layerize tab on the same page.
Write a prompt alone, or add up to 10 reference images to edit, combine, or restyle them.
Change one thing at a time. Areas you did not ask to change stay as they were, so details and colors hold up over many edits.
Ask for a transparent background and get a PNG with an alpha channel. No separate cut-out step.
The Layerize tab turns one finished image into a base image plus separate transparent layers you can move or replace.
Samples published by ByteDance Seed. Lighting, materials, and skin look more natural than in earlier Seedream versions.

A young father asleep on a sofa with a newborn on his chest, in warm window light.

A street candid with windblown hair, sunglasses, and a tilted frame, shot in hard sunlight.

Two diners in a European bistro on a summer evening, shot like a documentary photo.
It suits work that needs exact edits, clean cut-outs, or readable text.
Place a product into a new scene from two references, or retouch a label without touching the rest of the photo.
Lay out headlines, labels, and dense small text that stays legible, in English and other languages.
Export assets on a transparent background, ready to drop into a design.
Reuse a character from a reference image in new poses, outfits, and settings.
The Generate tab and the Layerize tab each have their own settings. The API accepts the same fields.
The prompt is required. Add up to 10 reference images (JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC, or HEIF, up to 30 MB each), or none for text to image.
Pick an aspect ratio (auto, 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 21:9, 2:1, or 1:2) and a resolution of 1K, 1.5K, or 2K. Through the API you can also request exact pixels.
Background can be auto, transparent, or opaque. Output is PNG or JPEG; a transparent background always returns PNG.
Upload one PNG or JPEG. Optionally name the elements to split, set the maximum number of images to return (2 to 100, base image included), and turn on text recognition for text layers.
Submit a task, then query the same external task ID until it finishes.
seedream_v5_5 for generation and editing, seedream_v5_5_layerize for layer splitting.
POST /api/openapi/submit with model, a unique out_task_id, and params such as prompt, image_url, aspect_ratio, resolution, background, and output_format.
POST /api/openapi/query with the same out_task_id until the status is completed or failed. Completed responses include the image URLs.
For generation, send size as "WIDTHxHEIGHT" instead of aspect_ratio and resolution. Layerize results return each layer with its z_index, bounding box, name, and description.
Compare Seedream 5.5 with other image models in the same playground.
| Control | Type | Default | Notes | |
|---|---|---|---|---|
| 01 | prompt | text | — | — |
| 02 | image_url | image | — | — |
| 03 | aspect_ratio | string | 1:1 | — |
| 04 | resolution | string | 2K | — |
| 05 | background | string | auto | — |
| 06 | output_format | string | png | — |
Describe the image, or write the edit as an instruction, such as “replace the vase with a bowl of lemons”.
Up to 10. Skip this step for text to image.
Choose an aspect ratio and resolution. Turn on a transparent background for stickers and cut-outs.
Download the result, or open the Layerize tab, upload it, and get a base image plus separate layers.
seedream_v5_5, seedream_v5_5_layerize
POST to /api/openapi/submit with the model, a unique out_task_id, and parameters.
POST to /api/openapi/query with the same out_task_id until the task completes.
Save outputs in your own storage rather than hot-linking them.
An image generation and editing model from ByteDance Seed, released in November 2026. It focuses on stable multi-round editing, native transparent backgrounds, layer splitting, and more natural lighting and textures.
Seedream 5.5 can make transparent backgrounds without a transparent input image, and its layer splitting returns up to 100 images instead of a base plus 16 layers. It also adds text recognition on text layers. Seedream 5.0 Pro keeps point-and-box selection editing.
Yes. Send the latest image back as the reference with your next instruction. Areas you do not mention are kept as they were, so the image does not drift over rounds.
Set Background to transparent. The result is a PNG with an alpha channel, including partial transparency for glass, smoke, or soft edges.
A base image plus separate transparent layers, up to the maximum you set. Each layer has a name, a description, and its position in the original image.
Generation is billed per image by resolution, and each reference image adds a small fee. Layerize is billed after it finishes, by the number of images returned. The live price table on this page is current, and failed jobs are not charged.
No. Resolution goes up to 2K. Through the API you can request exact pixel sizes within the model’s limits.
No. This page runs Seedream 5.5 through the AIReiter API with AIReiter credits. Use model ID seedream_v5_5 or seedream_v5_5_layerize.