Drop files here, paste files, or add a URL. JPEG, PNG or WebP · up to 10 MB each · up to 10 images
Output size: 1776×2368 px
Safety checks are always enabled in the playground and cannot be disabled.
Example Output
This is sample data. Generate a new result to see real output.
Pay as you go. No subscription, no hidden fees.
| Tier | 1K/1.5K | 2K |
|---|---|---|
| Generation & edit | $0.020/image 4 credits/image | $0.035/image 7 credits/image |
| Image to image | $0.020/image 4 credits/image | $0.035/image 7 credits/image |
| Tier | PoYo price |
|---|---|
| Reference image | $0.008/image 1.6 credits/image 10 images: $0.080 |
| Input image | $0.008/image 1.6 credits/image 10 images: $0.080 |
Charges are based on the final output.
ByteDance's image model for stable multi-turn editing, transparent PNGs, and image layers
Access the Seedream 5.5 API instantly on PoYo, no waitlist required. ByteDance Seed released Seedream 5.5 on November 9, 2026 as its next image creation model. Send a prompt to generate an image, or add up to 10 reference images in image_urls to edit them or combine them into one scene, and keep editing the same picture while the areas you leave out of the prompt stay intact.
Set background to transparent for a native RGBA PNG, choose one of 8 aspect ratios at 1K, 1.5K or 2K, or switch to seedream-5.5-layerize to split one image into a base image plus up to 99 transparent layers, each with its position, name, description and optional text OCR.
Stable multi-turn editing, native transparent output and fine-grained image layers from ByteDance's latest image model.
01
Repeated edits usually drift: colors shift, faces change and details you never asked to touch slowly disappear. Seedream 5.5 is built to preserve unedited regions at the pixel level and avoid color shifts there. Pass the previous output back in image_urls, say what to change and what to keep, and continue working on the same picture. In ByteDance's example, a portrait gains bubble gum, then sunglasses, then a different pair, while the hair, earrings and sweater stay the same. ByteDance demonstrates 100 consecutive editing rounds on one image.
02
Cutting a subject out of its background by hand loses soft edges, glass and shadows. Set background to transparent with PNG output and Seedream 5.5 returns an RGBA image with full or partial transparency, ready to place on any layout. It works from a prompt alone or from up to 10 reference images, and ordinary RGB images work as references.
03
A flat image is hard to rework once it is exported. Send it to seedream-5.5-layerize and get back a base image plus transparent PNG layers for the elements in the scene. Describe which elements to separate, or leave the prompt empty to split the main elements automatically. Each layer comes with its bounding box on the base image, so the layers stack back into the original layout. ByteDance's example splits an illustrated scene into a base image and 32 animal layers.
04
Localizing a poster or reusing a template means retyping every line of text. Turn on text_recognition and each regular text layer also returns its full text, the position of every line and approximate styles: the closest font family, color, bold, italic, underline, strikethrough, shadow and translucency. Your design tool can then change the words and place them back in the same spot.
05
Building one scene from several product photos, characters or style references usually takes manual compositing. Pass up to 10 images in image_urls, refer to them by their order in the prompt and say what each one contributes. Seedream 5.5 blends them into one image at 1K, 1.5K or 2K, and size auto follows the reference images. In ByteDance's example, two collage posters become one transparent sheet of separate stickers.
06
Generated photos often look plastic, and dense text tends to break apart. Seedream 5.5 renders more natural lighting, finer materials and more refined skin texture. In ByteDance's own evaluation, the severe text-collapse rate fell to 0.4% across Chinese, English and dense small text, which helps on posters, packaging and slides.
Turn flat posters and banners into editable layers with Seedream 5.5, then use the text OCR data to change copy, localize content and reuse the layout as a template inside your own editor.
Generate product cut-outs on transparent backgrounds, place them on new scenes, and refine the same hero image over several Seedream 5.5 edits without the packaging or logo drifting.
Create stickers, emojis, icons and overlays as transparent PNGs straight from a prompt or a photo, ready to drop into posts, chat apps and short videos.
Split illustrations and key art into separate character, prop and background layers with Seedream 5.5, so animators and level designers can move and reuse each element.
Revise posters and key visuals round after round while the approved parts stay untouched, and produce language versions by swapping the text on OCR-detected layers.
Create transparent prints, patterns and mockup assets for apparel, acrylic stands and packaging, then iterate on colors and details with Seedream 5.5 edits.
| Feature | Seedream 5.5 | Seedream 5.0 Pro | Seedream 4.5 | Seedream 4.0 |
|---|---|---|---|---|
| Best for | Repeated edits on one image, transparent PNG assets and layered, editable files | Dense layouts, multilingual text and region-precise edits | 4K commercial photography and sharp text rendering | Batches of up to 15 images from 1K to 4K |
| Model IDs | seedream-5.5, seedream-5.5-layerize | seedream-5.0-pro, seedream-5.0-pro-edit | seedream-4.5, seedream-4.5-edit | seedream-4, seedream-4-edit |
| Editing | Same model ID: add image_urls to edit | Separate edit model ID | Separate edit model ID | Separate edit model ID |
| Reference images | Up to 10 | Up to 10 | Up to 10 | Up to 10 (reference images plus outputs up to 15) |
| Maximum resolution | 2K (1K, 1.5K or 2K; up to 3136×1344) | 2K (1K or 2K) | 4K (2K or 4K presets, or a custom width and height) | 4K (1K, 2K or 4K) |
| Aspect ratios | 8 presets, plus auto for edits | 8 presets | 8 presets or a custom size | 8 presets |
| Images per request | 1 (layer decomposition: a base image plus up to 99 layers) | 1 | 1 to 15 | 1 to 15 |
| Transparent background (RGBA PNG) | Yes, from a prompt or from reference images | — | — | — |
| Layer decomposition | Base image plus up to 99 transparent layers, with positions, names and descriptions | — | — | — |
| Text-layer OCR | Text, line positions and approximate font, color and styles | — | — | — |
| Price on PoYo | 4 credits ($0.02) at 1K or 1.5K, 7 credits ($0.035) at 2K, plus 1.6 credits per reference image | 15 credits ($0.075) per image | 5 credits ($0.025) per image | 5 credits ($0.025) per image |
Step 1: Register & Create API Key
Sign in to PoYo and create an API key in the dashboard. The same key works for seedream-5.5, seedream-5.5-layerize and every other model on PoYo.
Create API Key →
Step 2: Top Up Credits
Add credits to your account. Each image costs 4 credits at 1K or 1.5K and 7 credits at 2K, plus 1.6 credits per reference or input image. Layer decomposition is billed per returned image, and the playground shows the estimate before you run.
Add Credits →
Step 3: Integrate the API
Send POST https://api.poyo.ai/api/generate/submit with "model": "seedream-5.5" and an input object with prompt, plus optional image_urls, size, resolution, background and output_format. For layers, use "model": "seedream-5.5-layerize" with one image in image_urls. Poll the returned task_id or pass a callback_url to receive the result.
View Documentation →
Call seedream-5.5 and seedream-5.5-layerize with the same request format and API key you use for every other image, video and audio model on PoYo.
See the per-image price for each resolution and the reference image add-on, and pay for layer decomposition by the images actually returned.
Try prompts, reference images, transparent backgrounds and layer settings in the browser, then copy the JSON request into your code.
Use documented request fields, task polling and callback URLs, and read layer positions and OCR data straight from the task result.
