Seedream 5.0 Lite API Guide: Text-to-Image and Image Editing

TL;DR

• Seedream 5.0 Lite is ByteDance's image model. On HiAPI it has two endpoints, seedream-5.0-lite/text-to-image and seedream-5.0-lite/image-to-image, both a flat $0.035 per image at 2K or 4K.
• It's strong at exact text in images and at targeted edits: change one object, swap a sign's wording, or composite up to 14 reference images while the rest of the scene stays put.
• aspect_ratio and resolution are required on every call. Open the text-to-image model page to try it; new accounts get 200 free credits.

Text-to-image or image-to-image

	Text-to-image	Image-to-image
Model id	seedream-5.0-lite/text-to-image	seedream-5.0-lite/image-to-image
Required	prompt, aspect_ratio, resolution	same, plus image_urls (1–14)
Price	$0.035 per image	$0.035 per image, however many references
Use it for	Posters, menus, packaging, infographics	Recoloring, text swaps, product-into-scene composites

Live HiAPI prices, checked October 19, 2026. Failed tasks aren't charged.

4K costs the same as 2K, so ask for 4K when you need print size. $1 buys about 28 images.

Text-to-image: text that comes out right

Put every word you want rendered in quotes and give each string a position:

A hand-drawn risograph-style poster for a fictional night market food festival, large Chinese calligraphy title '夜市食堂' at the top, English subtitle 'NIGHT MARKET CANTEEN' directly below it in a condensed sans-serif, a date line 'OCT 17-19 · RIVERSIDE PIER' at the bottom, warm lantern glow over crowded food stalls, steam rising from bamboo baskets, limited palette of vermilion, cream and ink blue, visible print grain, vintage poster composition

Generate this poster →

seedream-5.0-lite/text-to-image, 2K, $0.035.

The Chinese title, the English subtitle and the date line all came out as written. The same pattern works for menu boards, price tags and packaging; keep it under about six lines of text. Five more patterns, including a menu where every price is correct, are in Seedream 5.0 Lite prompt recipes.

Image-to-image: change one thing, keep the rest

The pattern: one sentence for the change, then name everything that must stay.

A local recolor. Before:

Change the matte white ceramic mug into a deep cobalt blue glazed mug with a glossy reflective finish, keep the steam, the oak desk, the open notebook, the fountain pen, the succulent, the lighting and the camera angle exactly unchanged, photorealistic

seedream-5.0-lite/image-to-image, one reference, 3:2, 2K, $0.035.

A text swap on a sign.

Replace the text on the wooden sign above the door so it reads 'BAKERY CORNER' in the same gold serif lettering style and size, keep the sign shape and wood texture, the brick facade, the glowing window, the olive trees and the overcast lighting exactly unchanged, photorealistic

seedream-5.0-lite/image-to-image, one reference, 3:2, 2K, $0.035.

A two-image composite. Refer to images by their order in image_urls:

Take the leather backpack from the first image and place it on the flat granite boulder in the campsite scene from the second image, leaning naturally at a slight angle, relit to match the warm golden-hour sunlight with a soft shadow falling across the rock, keep the stone fire ring, pine forest and mountain peaks unchanged, photorealistic composite

seedream-5.0-lite/image-to-image, two references, 3:2, 2K, $0.035.

Set aspect_ratio to match your input. If you ask for a different shape, part of the edit goes into re-cropping.

curl

curl -s -X POST https://api.hiapi.ai/v1/tasks \
  -H "Authorization: Bearer $HIAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-5.0-lite/text-to-image",
    "input": {
      "prompt": "A cafe chalkboard menu listing exactly three items: '\''FLAT WHITE 4.50'\'', '\''COLD BREW 5.00'\'', '\''CROISSANT 3.80'\''",
      "aspect_ratio": "3:4",
      "resolution": "2K"
    }
  }'

You get a taskId back. Poll GET https://api.hiapi.ai/v1/tasks/{taskId} until data.status is success, then download data.output[0].url. The link expires, so save the file right away.

For an edit, change the model to seedream-5.0-lite/image-to-image and add "image_urls": ["https://your-cdn.example.com/input.jpg"].

Python: generate, then edit

# pip install requests; export HIAPI_KEY=your-key first
import os, time, uuid, requests

API = "https://api.hiapi.ai/v1/tasks"
H = {"Authorization": f"Bearer {os.environ['HIAPI_KEY']}"}

def generate(model, timeout_s=300, **inp):
    body = {"model": model, "input": inp}
    # Reuse the same Idempotency-Key if you retry a submit, so a retry never bills twice
    r = requests.post(API, json=body, headers={**H, "Idempotency-Key": str(uuid.uuid4())}, timeout=30)
    res = r.json()
    if res.get("code") != 200:
        raise RuntimeError(res)  # 400s list every invalid field
    task_id = res["data"]["taskId"]
    deadline = time.time() + timeout_s
    while time.time() < deadline:
        d = requests.get(f"{API}/{task_id}", headers=H, timeout=30).json()["data"]
        if d["status"] == "success":
            return d["output"][0]["url"]
        if d["status"] == "fail":
            raise RuntimeError(d)
        time.sleep(5)
    raise TimeoutError(task_id)

# Generate a packshot
shot = generate("seedream-5.0-lite/text-to-image",
                prompt="Studio photo of a matte-black tea tin labeled 'HILLFOG' on slate, no other text",
                aspect_ratio="1:1", resolution="2K")

# Then make a colorway from it; the generated URL works as a reference
green = generate("seedream-5.0-lite/image-to-image",
                 prompt="Recolor the tin to deep forest green, keep the label, slate and lighting unchanged",
                 image_urls=[shot], aspect_ratio="1:1", resolution="2K")

with open("green.jpg", "wb") as f:
    f.write(requests.get(green).content)

Download each result as soon as it's ready if you plan to reuse it later; output links expire.

Parameters

Field	Required	Values	Notes
prompt	Yes	4–3,000 characters	Put rendered text in quotes
aspect_ratio	Yes	1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9	Match your input when editing
resolution	Yes	2K, 4K	Uppercase; same price
image_urls	Image-to-image only	1–14 URLs	jpeg, png or webp, up to 10 MB each

There's no seed, negative_prompt or output_format.

Errors you'll hit

All of these return HTTP 400 when you submit, so no task is created and nothing is billed:

• aspect_ratio: missing required field / resolution: missing required field: both are required, even though the docs list defaults.
• resolution: value must be one of '2K', '4K': there's no 1K. Use uppercase.
• image_urls: maxItems: got 15, want 14: 14 references at most.
• prompt: minLength: got 2, want 4: the prompt needs at least 4 characters.
• additional properties 'output_format' not allowed: the field doesn't exist on Lite.

Lite, Flash or Pro?

Model	Price per image	Pick it when
Seedream 5.0 Flash	$0.025	Volume and speed matter most
Seedream 5.0 Lite	$0.035 at 2K or 4K	Text in images, precise edits, 4K at no extra cost
Seedream 5.0 Pro	$0.05 (1K) / $0.10 (2K)	Highest detail for the final asset

A same-prompt test of the two cheaper tiers is in Seedream 5.0 Flash vs Lite. Full prices: Seedream 5.0 API pricing.

FAQ

Does each reference image cost extra?

No. Image-to-image is $0.035 per output, whether you send 1 reference or 14.

Can it render Chinese text?

Yes. The poster above has a Chinese title and an English subtitle in one image. Quote each string and anchor it to a position.

Will it keep my product exactly the same?

It keeps what you name. List every element that must stay (shape, label, lighting, camera angle); anything you leave out may change.

2K or 4K?

Same price. 2K has a long side of about 2,300 px and 4K about 4,096 px. 2K files are smaller and easier to review while you iterate; switch to 4K for print.