Nano Banana 2.1 text in images: posters and infographics on Sume

Treat it as a strong candidate, not a guarantee. fal's Nano Banana 2.1 page lists enhanced text rendering for posters and infographics among the model's features, and it supports 1K, 2K and 4K outputs. On Sume you can run it as google/nano-banana-2.1 for about $0.10 per 1K image, so a spelling test on ten candidate headlines costs about $1.00.

A cheap spelling test

No vendor page gives a guaranteed accuracy for your text, so run the words you will print. Keep headlines short, put the exact text in quotes, and compare the same prompt across rows. These rows list text-heavy use and are all on Sume.

Run the test

The script renders each headline on Nano Banana 2.1 and prints the URLs for a side-by-side check. It handles the 202 fallback by printing the job id instead of failing.

Read each picture for dropped letters, doubled letters and swapped words, and re-roll only the failures.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
for text in ["OPEN LATE FRIDAY", "Harvest Market 2026"]:
    r = requests.post(
        "https://api.sume.com/v1/images",
        headers=H,
        json={
            "model": "google/nano-banana-2.1",
            "prompt": f'poster, bold sans-serif headline exactly "{text}"',
            "aspect_ratio": "4:5",
        },
        timeout=60,
    )
    body = r.json()
    if r.status_code == 200:
        print(text, body["data"][0]["url"])
    else:
        print(text, r.status_code, body)

Where it fails

• Small body text and long paragraphs are still the riskiest.
• Fixed legal text should be added in your layout tool, not drawn.
• Lock the final wording before you spend on 4K.