How to Use Krea 3 API: First Image in 5 Minutes

TL;DR

• Krea 3 (krea-3/text-to-image) is a text-to-image model: one description yields photoreal shots, illustrations, or concept art, with prompt reasoning before generation and two variants — berry (quality, default) and jam (speed) — at the same price. Try it on the model page; new accounts get 200 free credits.
• Pricing is per image: Krea 3 at $0.036, with the cheaper family sibling Krea 2 Medium Turbo (krea-2-medium-turbo) at $0.0215 — full tiers on the pricing page.
• Everything runs on the HiAPI /v1/tasks endpoint: submit for a taskId, poll for the result, download from output[0].url.

What it does

• One line in, a full frame out. The model reasons about and expands your prompt before generating — a lighthouse still comes back with time of day, light, and composition; detailed prompts are followed as written.
• berry / jam, one model ID. Switch with variant; parameters and price are identical. Same seed gives different compositions and tones on each, so run both and pick.
• Six size presets up to 1K: square through widescreen (1024×1024 down to 1024×576), output as PNG or JPEG.
• Posters with headline text work too: quote the words; short headlines usually come out accurate.

Here's a real berry output:

Studio product photograph of a frosted glass perfume bottle with a brushed brass cap standing on a block of travertine, hard afternoon sunlight casting a long palm-leaf shadow across a warm beige wall, crisp reflections, high-end editorial look

krea-3/text-to-image, berry, square_hd, 1024×1024, $0.036.

Try this prompt →

Two models: which to use

	krea-3/text-to-image (Krea 3)	krea-2-medium-turbo (Krea 2 Medium Turbo)
Per image	$0.036	$0.0215
Size	6 presets (max 1024×1024), no custom sizes	aspect_ratio with 8 options incl. 2.35:1 cinema
Special parameters	variant (berry/jam), reasoning_level, negative_prompt (berry only)	creativity in 4 levels (raw/low/medium/high)
Prompt handling	Reasons and expands before generating; short lines still work	Renders as written — detail in, detail out
Best for	Final frames, fast exploration from short prompts, lettered posters	Volume output, widescreen framing, budget-sensitive daily work

Chain them: draft direction on Krea 2, finalize on Krea 3 — one draft plus one final is $0.0575. Scene-by-scene recipes are in the prompt guide.

Two ways to start

Browser, 3 steps: open the model page → paste the prompt and pick a size → generate. One image is $0.036.

API, 3 steps: POST /v1/tasks to submit → take the taskId → GET /v1/tasks/<taskId> until success, then download.

curl: your first image

curl -X POST https://api.hiapi.ai/v1/tasks \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: your-unique-key-001" \
  -d '{
  "model": "krea-3/text-to-image",
  "input": {
    "prompt": "Studio product photograph of a frosted glass perfume bottle with a brushed brass cap standing on a block of travertine, hard afternoon sunlight casting a long palm-leaf shadow across a warm beige wall, crisp reflections, high-end editorial look",
    "image_size": "square_hd",
    "variant": "berry",
    "reasoning_level": "medium"
  }
}'

Switching to Krea 2 Medium Turbo changes model and the input fields:

curl -X POST https://api.hiapi.ai/v1/tasks \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "krea-2-medium-turbo",
  "input": {
    "prompt": "Futuristic city skyline at dusk, neon reflections on wet streets, cinematic widescreen",
    "aspect_ratio": "2.35:1",
    "creativity": "low"
  }
}'

The response carries the taskId you'll poll with.

Python: submit, poll, download

import time, uuid, requests

BASE = "https://api.hiapi.ai/v1/tasks"
HEADERS = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
}

payload = {
    "model": "krea-3/text-to-image",
    "input": {
        "prompt": "Studio product photograph of a frosted glass perfume bottle with a brushed brass cap standing on a block of travertine, hard afternoon sunlight casting a long palm-leaf shadow across a warm beige wall, crisp reflections, high-end editorial look",
        "image_size": "square_hd",
        "variant": "berry",
        "reasoning_level": "medium",
    },
}

# Submit: the Idempotency-Key makes network retries safe
resp = requests.post(
    BASE,
    headers={**HEADERS, "Idempotency-Key": str(uuid.uuid4())},
    json=payload,
    timeout=30,
)
resp.raise_for_status()
task_id = resp.json()["taskId"]

# Poll: queued -> handling -> archiving -> success / fail
while True:
    task = requests.get(f"{BASE}/{task_id}", headers=HEADERS, timeout=30).json()
    print(task["status"])
    if task["status"] in ("success", "fail"):
        break
    time.sleep(3)

if task["status"] == "success":
    image_url = task["output"][0]["url"]   # temporary URL — download promptly
    img = requests.get(image_url, timeout=60)
    with open("output_image", "wb") as f:  # name the extension to match the actual format
        f.write(img.content)
    print("saved")

Parameters (krea-3/text-to-image)

Field	Required	Values	Notes
prompt	yes	1–5,000 chars	The frame you want; the more specific, the more literal
image_size	no	square_hd / square / portrait_4_3 / portrait_16_9 / landscape_4_3 / landscape_16_9, default square_hd	Size presets up to 1024×1024; no auto or custom sizes
variant	no	berry / jam, default berry	berry favors quality, jam favors speed; same price
reasoning_level	no	low / medium, default medium	How much the prompt is expanded; medium for short prompts, low for detailed ones
negative_prompt	no	≤1,024 chars	What to keep out; works on berry only
output_format	no	png / jpeg, default png	Output format
seed	no	integer 0–2147483647	Reproducible within a variant; berry and jam don't cross-reproduce

Krea 2 Medium Turbo's fields differ: prompt (required), aspect_ratio (8 options, default 1:1), creativity (raw/low/medium/high, default low), seed. Any field outside the table returns a 400.

Errors you'll actually hit (verbatim)

What you did	Response
Sent auto as image_size	400 invalid input: image_size: value must be one of 'square_hd', 'square', 'portrait_4_3', 'portrait_16_9', 'landscape_4_3', 'landscape_16_9'
Sent an unknown field (e.g. num_images)	400 invalid input: <root>: additional properties 'num_images' not allowed
Sent an empty prompt	400 invalid input: prompt: minLength: got 0, want 1
Misspelled variant	400 invalid input: variant: value must be one of 'berry', 'jam'
Out of credits	402 insufficient account balance; top up and retry

Requests that fail parameter validation are not charged. Re-submitting the same Idempotency-Key returns 409 (processing) or 422 (payload mismatch); use a fresh key.

Before you ship

• Callbacks: pass "callback": {"url": "https://..."} at the request top level and HiAPI notifies your service at the terminal state.
• Persist outputs: output[0].url is temporary — download it into your own storage right away.
• Six sizes only: for 2.35:1 cinema or more ratios, use Krea 2 Medium Turbo's aspect_ratio.
• Seeds reproduce within a variant: berry and jam on the same seed give different images; fix variant + seed for series work.

FAQ

Do short prompts work?

Yes. The model expands them — a lighthouse returns a complete frame; write in detail when you want control.

Are berry and jam just fast and slow?

No. Same seed gives different compositions and tones at comparable quality — run both and pick.

Does negative_prompt work on both variants?

Only on berry; sending it with jam does nothing.

How many images per call?

One. Submit more tasks or rerun with different seeds.

How far does $10 go?

About 277 images at $0.036 each. New accounts get 200 credits (2,000 credits = $1).

Next: take the perfume prompt and generate a test image, then copy scene templates from the prompt guide.