AI Image Resolution API — 1K, 2K, and 4K Generation Pricing Explained

A developer building a print-on-demand storefront generates a product mockup at the default resolution, sends it to the print vendor, and gets a rejection — the file's too low-res for a 12x16 canvas print. The fix isn't a different model, it's a different resolution parameter, and it changes the per-image cost. Resolution tiers are one of the few places where an AI image API's pricing isn't flat per call — picking the wrong one either wastes money on 4K thumbnails or ships unprintable 1K files.

This is a tutorial: what resolution options GenRelay's image models expose, what each tier costs, and how to request the right one for a given output target.

What Resolution Options Does the GenRelay Image API Support?

Resolution support depends on the model — Nano Banana Pro and Nano Banana 2 expose selectable tiers, GPT-image-2 generates at a fixed output resolution.

Model	1K	2K	4K
Nano Banana Pro	Supported	Supported	Supported
Nano Banana 2	Supported	—	Supported
GPT-image-2	Fixed output (no tier param)	—	—

Definition: a resolution tier is a request parameter that sets the pixel dimensions of the generated image — 1K (roughly 1024px on the long edge), 2K (roughly 2048px), or 4K (roughly 3840–4096px) — independent of aspect ratio, which is a separate parameter covered in the aspect ratio API guide.

How Much Does Each Resolution Tier Cost?

Higher resolution costs more per image for the two models that support tiers, but not proportionally — Nano Banana Pro's 1K and 2K are priced identically.

Model	1K	2K	4K
Nano Banana Pro	$0.030	$0.030	$0.042
Nano Banana 2	$0.020	—	$0.036
GPT-image-2	$0.014 (flat)	—	—

For a batch of 500 product images generated at 2K with Nano Banana Pro: 500 × $0.030 = $15. Bumping the same batch to 4K for a print catalog: 500 × $0.042 = $21 — a $6 difference for a 4x pixel-count increase, which is cheap enough that the decision usually comes down to whether the output actually needs 4K, not whether the API cost justifies it.

How Do I Request a Specific Resolution via API?

Pass the resolution parameter with the job request — omit it and the model falls back to its default tier (1K for both Nano Banana models).

import requests

API_KEY = "YOUR_GENRELAY_KEY"
BASE_URL = "https://genrelay.ai/v1"

def generate_image(prompt, model="nano-banana-pro", resolution="2k"):
    response = requests.post(
        f"{BASE_URL}/images/generations",
        headers={
            "Authorization": f"Bearer {API_KEY}",
            "Content-Type": "application/json"
        },
        json={
            "model": model,
            "prompt": prompt,
            "resolution": resolution
        }
    )
    return response.json()

# Web thumbnail — 1K is sufficient, cheapest tier
thumbnail = generate_image(
    "minimalist product photo of a ceramic mug on a white background",
    resolution="1k"
)

# Print catalog asset — 4K for a large-format print
print_asset = generate_image(
    "minimalist product photo of a ceramic mug on a white background",
    model="nano-banana-pro",
    resolution="4k"
)

Requesting resolution="4k" against GPT-image-2 has no effect — the parameter is silently ignored and the image returns at its fixed output size — so route 4K-dependent jobs to Nano Banana Pro or Nano Banana 2 at the request-building stage rather than discovering the mismatch after the file fails a print check.

def pick_model_for_target(resolution):
    if resolution == "4k":
        return "nano-banana-pro"  # or "nano-banana-2"
    return "gpt-image-2"  # cheapest flat-rate option when tier doesn't matter

Which Resolution Should I Use for Which Use Case?

Match the tier to the output's actual display size, not to "higher is always better" — 4K files are larger, slower to generate, and cost more per image for no visible gain on a phone screen.

• Social media posts, in-app thumbnails, chat previews: 1K. Screens rarely render above 1080–1440px, so 2K or 4K pixels are discarded on display.
• Website hero images, blog headers, e-commerce listing photos: 2K. Enough headroom for retina displays and moderate cropping without visible upscaling artifacts.
• Print catalogs, large-format marketing assets, poster-sized output: 4K. Print resolution requirements (300 DPI at physical size) generally exceed what 1K or 2K can supply without visible softness.
• High-volume, resolution-agnostic use cases (batch icon sets, placeholder art, A/B test variants): GPT-image-2's flat $0.014/image rate, since resolution tiering doesn't apply and the per-image cost is the lowest of the three models.

FAQ

Does GPT-image-2 support a 4K output option?
No. GPT-image-2 generates at a fixed resolution regardless of request parameters. Use Nano Banana Pro or Nano Banana 2 when a job specifically needs 4K output.

Is there a quality difference between Nano Banana Pro's 1K and 2K tiers beyond pixel count?
Both are priced identically ($0.030), and the underlying generation quality is the same — 2K simply renders at higher pixel dimensions. There's no cost penalty for defaulting to 2K over 1K on Nano Banana Pro specifically.

What happens if I request a resolution a model doesn't support?
Nano Banana 2 only supports 1K and 4K — requesting 2k against it returns a validation error rather than silently rounding to the nearest tier. Check the model's supported tiers before submitting a batch job to avoid a partial batch failure.

Can I mix resolution tiers within a single batch request?
Each generation call takes one resolution value. To generate a mixed batch (some 1K, some 4K), submit separate calls per tier rather than one batch call with per-image overrides — see the batch image generation guide for concurrency patterns.

Is there a free tier to test resolution output before committing to a 4K batch?
Yes. GenRelay includes free credits on signup, enough to generate sample images at each tier and inspect the actual pixel output before budgeting a full print-resolution batch.

---

As of September 2026. Pricing subject to change — verify current rates at genrelay.ai.