Seedream 4.5 API — Access via an AI API Gateway (Jimeng · ByteDance Text-to-Image)

Generate images with ByteDance's Seedream 4.5 (seedream-4.5) through byesu — an AI API gateway with OpenAI-compatible and Anthropic-native endpoints, one key for Claude / GPT-5.6 / Grok / Gemini and media models, pay-as-you-go with no subscription. Seedream 4.5 combines strong scene understanding with accurate instruction following — it excels at posters, e-commerce imagery, complex multi-subject compositions, and bilingual (English / Chinese) prompts. Billing is per image at $0.05 / image, supporting 2K / 4K resolution and up to 6 reference images.

byesu routes requests to the official upstream (not reverse-engineered, not mirrored, not cracked), with transparent pay-as-you-go billing. A single key lets you call Seedream 4.5 alongside other mainstream text-to-image / video models.

• Base URL: https://byesu.com/v1
• Auth header: Authorization: Bearer sk-your-token
• Text-to-image endpoint: POST /v1/images/generations (returns image base64 synchronously)
• Image-to-image endpoint: POST /v1/images/edits (multipart; the image field accepts multiple files, each ≤8MB)

The token must use the media group

Tokens for image models must belong to the media / media-gen group, otherwise you'll get a "No available channel" error. Go to Console → Tokens and create a token in that group. See Choosing a Group.

Seedream 4.5 Pricing

Item	Value
Model name	seedream-4.5
Billing	Per image
Price	$0.05 / image
Resolution	2K · 4K
Reference image limit	≤ 6 images (each ≤8MB)

Official upstream routing — no quality downgrade. Live prices show up directly on the "Generate" button when you pick a model / resolution in the Generation Playground; the console also shows your live billed price side by side with the official list price.

How to Call

Text-to-Image (curl)

curl https://byesu.com/v1/images/generations \
  -H "Authorization: Bearer sk-your-token" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-4.5",
    "prompt": "Premium e-commerce hero shot, skincare bottle with blue-and-white porcelain pattern, soft studio lighting, centered composition, generous negative space, minimal luxury style",
    "size": "2048x2048"
  }'

Response (the image is base64 — read data[0].b64_json):

{ "data": [ { "b64_json": "<base64 string>" } ] }

size sets both aspect ratio and resolution

size takes "width x height" in pixels and determines both the aspect ratio and the resolution tier: long side ≈2048 → 2K, ≈3840 → 4K. E.g. "2048x1152" = 16:9 at 2K, "3840x2160" = 16:9 at 4K.

Text-to-Image (Python)

import requests, base64

resp = requests.post(
    "https://byesu.com/v1/images/generations",
    headers={"Authorization": "Bearer sk-your-token"},
    json={
        "model": "seedream-4.5",
        "prompt": "Premium e-commerce hero shot, skincare bottle with blue-and-white porcelain pattern, soft studio lighting, centered composition, generous negative space, minimal luxury style",
        "size": "2048x2048",
    },
).json()

b64 = resp["data"][0]["b64_json"]
open("out.png", "wb").write(base64.b64decode(b64))

Image-to-Image / Multiple Reference Images (curl)

When passing reference images, switch to POST /v1/images/edits with multipart/form-data, uploading each reference image as an image file field (repeatable — Seedream 4.5 accepts up to 6 images, each ≤8MB):

curl https://byesu.com/v1/images/edits \
  -H "Authorization: Bearer sk-your-token" \
  -F model="seedream-4.5" \
  -F prompt="Keep the person unchanged, replace the background with a warm sunset rooftop terrace" \
  -F size="2048x2048" \
  -F image=@ref1.png \
  -F image=@ref2.png

The response format is the same as text-to-image. If something goes wrong, see Common Errors.

Key Parameters

Field	Required	Description
model	✅	Always seedream-4.5
prompt	✅	Prompt in English or Chinese; being specific (subject + scene + style + lighting) yields better results
size		"width x height" in pixels — determines both the aspect ratio and the resolution tier; long side ≈2048 → 2K, ≈3840 → 4K. Defaults to 1:1 at the base tier if omitted
image	Required for image-to-image	/v1/images/edits only; reference image file field, repeatable, Seedream 4.5 ≤6 images, each ≤8MB

Common aspect ratios 1:1 · 16:9 · 9:16 · 4:3 · 3:4 — just express them via size.

Why Use byesu for Seedream 4.5

• Official upstream, no quality downgrade — requests reach the real Seedream 4.5, not a reverse-engineered or mirrored stand-in; image quality matches the official service.
• One gateway, many models — the same key also covers Claude, GPT-5.6, Grok, Gemini and other text / image / video models; switch by changing one string.
• Transparent pricing — billed per image at $0.05 / image; the console shows your live billed price side by side with the official list price.
• Auto-refund on failure — failed media generation tasks are refunded automatically, so you never pay for failed results.
• Multiple reference images — supports up to 6 reference images, ideal for image-to-image workflows like background swaps, multi-image compositing, and style transfer.
• Pay-as-you-go, no subscription — top up with USDT, Alipay or WeChat Pay and pay only for what you generate.

Related Links

• Image / Video Generation Overview — all media models and the playground
• Image Generation Model List — Seedream 4.5 and other text-to-image models
• Calling via API (Images / Video) — full API format and code
• Quickstart (3 steps) — get a token + set the URL + pick a model
• Console Registration / Create a Token — create a media / media-gen group token
• Online Generation Playground — no-code image generation, see live pricing

FAQ

How do I get Seedream 4.5 API access?

Sign up at the byesu console, create a token in the media / media-gen group, and call seedream-4.5 through the OpenAI-compatible endpoint POST https://byesu.com/v1/images/generations with the auth header Authorization: Bearer sk-your-token. byesu is an AI API gateway — one key also covers Claude, GPT-5.6, Grok, Gemini and other image / video models, billed pay-as-you-go with no subscription. Requests route to the official upstream, so there is no quality downgrade.

How much does the seedream-4.5 API cost?

Billing is per image, at $0.05 per image. Resolution supports 2K / 4K, and up to 6 reference images. Live pricing appears on the Generate button when you pick a model / resolution in the Generation Playground, and the console also shows your live billed price side by side with the official list price. Failed media tasks are refunded automatically.

How do I call the Seedream API? Which endpoint do I use?

For text-to-image, use POST https://byesu.com/v1/images/generations (JSON, returns image base64 synchronously — read data[0].b64_json); for image-to-image with reference images, use POST /v1/images/edits (multipart, the image field accepts multiple files, each ≤8MB). Set model to seedream-4.5 and the auth header to Authorization: Bearer sk-your-token.

How many reference images does Jimeng / Seedream 4.5 support, and what is the maximum resolution?

Seedream 4.5 supports up to 6 reference images (each ≤8MB), with resolution up to 2K and 4K. Reference images go through the /v1/images/edits endpoint, uploaded repeatedly via the image file field — ideal for image-to-image workflows like background swaps, multi-image compositing, and style transfer.

How do I use seedream-4.5 in Claude Code or a client app?

Seedream 4.5 is an image generation model, called via the REST endpoint /v1/images/generations. The simplest option is the online Generation Playground (no-code — just enter your token and generate) or sending HTTP requests from your own program. Chat/coding clients like Claude Code / Cherry Studio mainly use the text endpoint /v1/chat/completions and generally do not call image endpoints directly; to integrate image generation into an app, wire it up using the curl / Python examples on this page.

What does a "No available channel" error mean when calling seedream-4.5?

It usually means the token was created in the wrong group. Tokens for image / video models must belong to the media / media-gen group. Go to Console → Tokens and create a new token in that group, then try again. For other errors (401, insufficient balance, exceeding the reference-image limit), see the Common Errors page in the docs.