ByteDance Seedream 5 Pro: Pricing and API Code (2026)

Seedream 5 Pro is ByteDance’s highest-quality Seedream image model, built for finished work: product shots, illustration and anything a client will actually see. This guide explains what it does, how it compares with Seedream 5 Lite and other image models, what each image costs, and the exact API calls to generate, edit and split images into layers. It ends with the mistakes that waste the most credits.

What is Seedream 5 Pro?

Seedream is ByteDance’s family of text-to-image models. A text-to-image model takes a natural-language description, the prompt, and produces a matching picture. Most current models, Seedream included, also do image-to-image work: you supply one or more existing pictures, called reference images, and the model uses them to keep a product, character or style consistent in the new output.

Seedream 5 Pro is the top tier of the Seedream 5 generation. It was released on 8 July 2026. Its job is quality rather than speed: polished illustration and product shots at 1K or 2K resolution, with up to 10 reference images per request. It supports eight aspect ratios: 16:9, 1:1, 9:16, 21:9, 4:3, 3:4, 3:2 and 2:3. That covers widescreen banners, square social posts, vertical stories, ultra-wide hero images and the usual print proportions.

Two related models are worth knowing about:

• Seedream 5 Lite, released on 25 February 2026, is the fast everyday tier. It outputs at 2K or 3K and accepts up to 14 reference images.
• Seedream 5 Pro Layer Decomposition was released the same day as Pro. It takes an existing image and separates it into layers, so you can edit the background, subject and text on their own.

Naming varies from site to site. “Seedream 5.0 Pro”, “ByteDance Seedream 5 Pro” and “Seedream 5 Pro” all refer to the same model. If you want the history of the whole 5.0 family, including the Flash tier, our Seedream 5.0 overview covers it. This article stays focused on the Pro tier and how to use it well.

Seedream 5 Pro vs Seedream 5 Lite and other image models

The most common question is whether Pro is worth it over Lite. The short answer: Pro is about how finished the image looks, not how many pixels it has. Lite actually goes to a higher resolution (3K) and accepts more reference images (14 vs 10). What you pay extra for with Pro is finish quality on the images that matter.

Here is how Seedream 5 Pro sits next to the other image models you can call with the same API key:

Model	Maker	Output size	Reference images	Credits per image	Best for
Seedream 5 Pro	ByteDance	1K or 2K	Up to 10	7 (1K), 14 (2K)	Polished illustration and product shots
Seedream 5 Lite	ByteDance	2K or 3K	Up to 14	5.5 (rounded up to 6 for a single image)	Fast everyday generation, higher resolution
Nano Banana 2 Lite	Google	1K	Up to 10	4	Cheap, fast images and edits
GPT Image 1.5	OpenAI	Low, medium or high quality	Up to 4	3, 4 or 22	Prompt adherence, legible text, transparent backgrounds
Flux 3 Klein Text to Image	Black Forest Labs	1K, 2K or 4K	None (text only)	1.6, 2.5 or 6.5	Very cheap drafts, 4K output

When to choose Pro

• The image is the final deliverable: a product page hero, an ad creative, a cover illustration.
• You need product consistency across a set and have good reference photos to feed in.
• You need 3:4 portrait framing. Pro supports it and Lite does not.

When something else is the better fit

• Exploration and volume: Lite, Nano Banana 2 Lite or Flux 3 Klein at 1K cost less, so you can try many ideas before you commit.
• Transparent backgrounds: GPT Image 1.5 has a background option. Pro does not list one.
• 4K output: Flux 3 Klein offers 4K. Pro tops out at 2K.

Many people search for “Seedream 5 vs Nano Banana Pro”. Neither model wins on every prompt, and published side-by-side tests rarely match your own subject matter. The reliable method is to run your real prompts, with your real reference images, through both models and compare. Our Nano Banana API guide covers the Google side of that test.

How much does Seedream 5 Pro cost?

On Unicorn API, Seedream 5 Pro is priced in credits per output image:

• 1K: 7 credits per image
• 2K: 14 credits per image

Credits come from a monthly plan or from top-ups. Plan credits cost $0.01 each: Starter is $25, Creator $50, Studio $100 and Pro $200 per month. Top-up credits cost $0.015 each and never expire. In dollars, that means:

Resolution	Credits	On a monthly plan	With top-up credits
1K	7	$0.07	$0.105
2K	14	$0.14	$0.21

The Starter plan’s 2,500 credits buy about 357 Seedream 5 Pro images at 1K, or 178 at 2K. Full plan details are on the pricing page.

Three billing details are easy to miss:

1. The n field multiplies the price. Asking for four images with "n": 4 at 2K costs 56 credits.
2. Prices round up to whole credits. That doesn’t affect Pro, but it does affect Lite: one Lite image at 5.5 credits is billed as 6, while two images (n=2) cost exactly 11.
3. You can check the price before running anything. POST /v1/models/price returns the exact cost of a request, which is useful before large batches. The request format is in the API documentation.

The “Seedream 5 free online” sites that rank for this keyword usually come with daily caps, queues or unclear licence terms. They are fine for a quick look. For anything you plan to ship, paying a known amount per image through an API is simpler to budget and to automate.

How to call the Seedream 5 Pro API

Image models on Unicorn API run as jobs. A job is a piece of work that runs in the background: you submit it, get back an id, and poll until it finishes. That pattern copes with generation times of several seconds or more without holding an HTTP connection open.

The steps:

1. Create an API key on the keys page and export it as UNICORN_API_KEY.
2. Send POST /v1/jobs with "model": "seedream-5-pro" and an input object.
3. Poll GET /v1/jobs/{id} until status is "succeeded".
4. Read the image URL from outputs[0].url.

curl

curl -X POST https://api.unicornapi.net/v1/jobs \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-5-pro",
    "input": {
      "prompt": "A frosted glass perfume bottle on wet black slate, single soft key light from the left, shallow depth of field, studio product photography",
      "resolution": "2K",
      "aspect_ratio": "4:3"
    }
  }'

# Then poll with the id from the response:
curl https://api.unicornapi.net/v1/jobs/JOB_ID \
  -H "Authorization: Bearer $UNICORN_API_KEY"

Python

This script submits a job, polls with a timeout and downloads the result:

import os
import time
import requests

API = "https://api.unicornapi.net"
HEADERS = {
    "Authorization": f"Bearer {os.environ['UNICORN_API_KEY']}",
    "Content-Type": "application/json",
}


def generate(prompt, resolution="1K", aspect_ratio="1:1", ref_images=None, n=1):
    payload = {
        "model": "seedream-5-pro",
        "input": {
            "prompt": prompt,
            "resolution": resolution,
            "aspect_ratio": aspect_ratio,
            "n": n,
        },
    }
    if ref_images:
        payload["input"]["ref_images"] = ref_images

    resp = requests.post(f"{API}/v1/jobs", headers=HEADERS, json=payload)
    resp.raise_for_status()
    job_id = resp.json()["id"]

    deadline = time.time() + 300
    while time.time() < deadline:
        job = requests.get(f"{API}/v1/jobs/{job_id}", headers=HEADERS).json()
        if job["status"] == "succeeded":
            return [o["url"] for o in job["outputs"]]
        if job["status"] == "failed":
            raise RuntimeError(f"Job {job_id} failed: {job}")
        time.sleep(3)
    raise TimeoutError(f"Job {job_id} did not finish in time")


if __name__ == "__main__":
    urls = generate(
        "Flat vector illustration of a lighthouse at dusk, limited palette of teal and coral",
        resolution="2K",
        aspect_ratio="16:9",
    )
    for i, url in enumerate(urls):
        with open(f"seedream_{i}.png", "wb") as f:
            f.write(requests.get(url).content)
        print("saved", url)

If you would rather try prompts before writing code, the image playground runs the same models in the browser. If you work in Cursor or Claude Code, the MCP setup lets your agent call Seedream 5 Pro directly.

Editing with reference images and layer decomposition

Text-only generation is only half of what Seedream 5 Pro is useful for. The other half is editing and consistency, and that runs through the ref_images field.

Reference images

Pass up to 10 image URLs in ref_images, then describe in the prompt what to keep and what to change. Typical uses:

• Product consistency: two or three clean photos of the product from different angles, plus a prompt for the new scene.
• Style transfer: one or two images in the target style, plus a prompt for new content in that style.
• Direct edits: one source image with a prompt such as “replace the background with a sunlit marble countertop, keep the bottle unchanged”.

{
  "model": "seedream-5-pro",
  "input": {
    "prompt": "Place this exact bottle on a sunlit marble countertop with a sprig of lavender, keep label text and bottle shape unchanged",
    "ref_images": [
      "https://example.com/bottle-front.png",
      "https://example.com/bottle-side.png"
    ],
    "resolution": "2K",
    "aspect_ratio": "3:4"
  }
}

Be explicit about what each reference is for. “Use the bottle from the references” works better than leaving the model to guess whether you want the bottle, the lighting or the background.

Layer decomposition

Seedream 5 Pro Layer Decomposition works the other way round. It starts from a finished image (image_url) plus a prompt and returns the image split into separate layers. That helps when a designer needs to move a product, swap a background or fix a headline without regenerating the whole picture. The size field accepts 1K, 1.5K, 2K or "auto":

{
  "model": "seedream-5-pro-layer-decomposition",
  "input": {
    "size": "auto",
    "prompt": "Separate the product, the background and the headline text",
    "image_url": "https://example.com/poster.png"
  }
}

Pricing is 7 credits per output at 1K or 1.5K and 14 at 2K. Because billing is per output, a request that returns several layers may cost more than a single image. Check the price endpoint before running large batches.

For simple edits where Pro-level finish isn’t needed, Flux 3 Klein Edit Image is far cheaper at 1.6 credits per 1K image. It takes source images in image_urls.

Worked example: a product shot from draft to final

Here is a realistic workflow for a single e-commerce hero image, with the cost of each step.

1. Write a structured prompt

Seedream 5 Pro responds well to prompts that read like a photographer’s brief. Cover five things in this order: subject, setting, lighting, camera and finish.

A matte ceramic coffee mug in sage green, on a light oak table beside a folded linen napkin, morning window light from the right with soft shadows, 50mm lens at eye level with shallow depth of field, clean commercial product photography, no text.

Adding “no text” matters on product shots. Without it, models sometimes invent labels or logos.

2. Explore at 1K

Run the prompt with "resolution": "1K" and "n": 4. That costs 28 credits ($0.28 on a plan) and gives you four compositions to choose from. If none work, adjust the prompt and run again. Exploring at 1K costs half as much as exploring at 2K.

3. Lock in the product with references

Once you like a composition, add two real photos of the mug in ref_images so the colour and handle shape match the actual product. Adjust the prompt to say “this exact mug from the reference images”.

4. Render the final at 2K

Run once with "resolution": "2K", "n": 1 and the aspect ratio your page layout needs, for example 4:3. That costs 14 credits.

Total: 42 credits, or $0.42 on a monthly plan, for four drafts and one final 2K hero image. Scale that to a catalogue of 50 products and you are at 2,100 credits, which fits inside the Starter plan.

5. Store the output

Download the file from outputs[0].url and save it in your own storage or CDN. Do not hotlink generated URLs from a production page.

Common mistakes with Seedream 5 Pro

Most wasted credits and disappointing results come from a short list of habits:

• Exploring at 2K. Every 2K draft costs double. Find the composition at 1K, then render the keeper at 2K.
• Using Pro for volume work. If you need hundreds of social variants or quick concept art, Lite or Flux 3 Klein will cover it for less. Save Seedream 5 Pro for images that ship.
• Expecting 4K. Pro outputs 1K or 2K. If you need more pixels, use Lite at 3K, Flux 3 Klein at 4K, or upscale a 2K Pro image afterwards.
• Forgetting that n multiplies the price. "n": 4 at 2K is 56 credits, not 14. Check with the price endpoint when you build batch jobs.
• Feeding in poor or conflicting references. Ten references with different lighting, angles and backgrounds pull the model in different directions. Two or three clean, consistent photos usually beat ten mixed ones.
• Vague prompts. “A nice product photo of a mug” leaves lighting, angle and setting to chance. The five-part brief above removes most of the guesswork.
• Using an aspect ratio the model doesn’t support. Pro supports 3:4 and Lite does not. If you switch models in code, validate the aspect ratio for each one.
• Polling without a timeout. Always cap your polling loop and handle a failed status, as the Python example does, so one stuck job can’t hang a batch.

A quick note on the “CapCut Seedream 5” searches: ByteDance’s consumer creative apps are a fine way to try the model by hand. They are built for one person clicking through edits, not for generating a product catalogue from a script. When you need repeatable output, cost control and code, an API is the practical route. Seedream 5 Pro is one model id on a key that also covers every other model in the model catalog.