Seedream 5.0 Pro
Generate, transform, reference, and decompose images with Seedream 5.0 Pro.
The public model ID is seedream-5-0-pro. It provides four image capabilities through the asynchronous media task API. Clients use only the public model ID and must not send or depend on an internal upstream model name.
Model information
| Item | Value |
|---|---|
| Model ID | seedream-5-0-pro |
| Endpoint | POST /v1/router/task/create |
| Task types | Txt2Img, Img2Img, Multi-reference, LayerDecomposition |
| Channel | Official API Relay |
Authorize the model for your API key first. After creation, save the returned tsk_* task ID and poll GET /v1/router/tasks/{task_id} until the status becomes 2, 3, or 4.
Request parameters
| Parameter | Type | Required for | Description |
|---|---|---|---|
task | string | All requests | One of the four task types |
model | string | All requests | Fixed to seedream-5-0-pro |
input.prompt | string | Generation | Generation or editing prompt |
input.image | string | Single image | Public HTTPS image URL for Img2Img and LayerDecomposition |
input.images | string[] | References | 2–10 public HTTPS image URLs for Multi-reference |
input.size | string | All requests | 1K, 1.5K, or 2K; layer decomposition additionally supports auto |
callback_url | string | No | Optional public HTTPS webhook. See Media task callbacks. |
Txt2Img accepts no input image. Img2Img and LayerDecomposition each require exactly one. Do not send both input.image and input.images.
Fixed prices
These official credit prices apply before coefficients. The gateway reserves the amount derived from the request when creating a task and settles it only after success. Failure or cancellation releases the reservation without a charge.
| Task | 1K | 1.5K | 2K |
|---|---|---|---|
| Text / single-image generation | 4.5 credits per image | 4.5 credits per image | 9 credits per image |
| Multi-reference | 4.5 + 0.3 × (N−1) | 4.5 + 0.3 × (N−1) | 9 + 0.3 × (N−1) |
| Layer decomposition | 2.25 credits per actual layer | 2.25 credits per actual layer | 4.5 credits per actual layer |
For multi-reference generation, N is the number of input images. Layer decomposition reserves the maximum of 16 layers at creation and settles 1–16 layers actually returned; the Base Image is not counted as a layer. With size=auto, each layer up to 2,610,000 pixels costs 2.25 credits, and a larger layer costs 4.5 credits.
Request examples
# Multi-reference generation
curl https://hopsapi.com/v1/router/task/create \
-H "Authorization: Bearer sk-YOUR_KEY" \
-H "content-type: application/json" \
-d '{"task":"Multi-reference","model":"seedream-5-0-pro","input":{"prompt":"keep the character consistent","images":["https://example.com/a.png","https://example.com/b.png"],"size":"1.5K"}}'# Layer decomposition
curl https://hopsapi.com/v1/router/task/create \
-H "Authorization: Bearer sk-YOUR_KEY" \
-H "content-type: application/json" \
-d '{"task":"LayerDecomposition","model":"seedream-5-0-pro","input":{"image":"https://example.com/source.png","size":"2K"}}'Invalid parameters return HTTP 400 without submitting an upstream task. An unavailable or unauthorized model returns HTTP 403. If a terminal result does not contain a valid layer count or verifiable dimensions for auto, the gateway withholds the result and never settles it as a zero-cost task.