Translate packaging text: Ideogram 4.5 vs Nano Banana Pro per language

Translate label text in one image per language. On Sume Ideogram 4.5 costs $0.0375 to $0.275 per edit; Nano Banana Pro $0.1875 at 1K to 2K. Limits compared.

5 min readSume
All posts

For translating packaging text one language at a time, Ideogram 4.5 is the cheaper route on Sume at low ($0.0375) and medium ($0.075) quality, and Nano Banana Pro is cheaper only against Ideogram high ($0.275). Nano Banana Pro bills $0.1875 per image at 1K and 2K and $0.375 at 4K, so six languages cost $1.125 at 1K or 2K against $0.225 for Ideogram at low.

Both take the packaging render as a reference and return a new image through POST /v1/images. Price is not the whole decision, so this post lines up the limits that change what you can ship, from Sume's docs and catalog (read 2026-10-05). It makes no claim about which one spells Spanish more accurately; that needs your artwork.

What does each model take?

The prices are Sume's list x 1.25. Ideogram's list is $0.03, $0.06 and $0.22 by quality, and Nano Banana Pro's is $0.15 at 1K and 2K and $0.30 at 4K.

Ideogram 4.5 and Nano Banana Pro on Sume. Image API docs and Sume catalog, read 2026-10-05.
FieldIdeogram 4.5Nano Banana Pro
Public idideogram/ideogram-v4.5google/nano-banana-pro
Price per edit at Sume$0.0375 / $0.075 / $0.275 by quality$0.1875 (1K, 2K), $0.375 (4K)
Price depends onquality, not sizeresolution tier
Resolution tiers1K, 2K0.5K, 1K, 2K, 4K
Reference images5 (first is edited)up to 10
Aspect ratios15, from 1:3 to 3:111 incl. auto, 4:5, 21:9 and 9:16
Edit without aspect_ratiokeeps the source shapesee the ratio note below
mask_url400400

What is the cost per language?

A translated pack set is one edit per language, from one master render. Here is the bill for the six most common retail languages and for 12.

One edit per language at Sume list x 1.25; Nano Banana Pro at 1K or 2K. Docs read 2026-10-05.
Setup6 languages12 languages
Ideogram 4.5 low$0.225$0.45
Ideogram 4.5 medium$0.45$0.90
Ideogram 4.5 high$1.65$3.30
Nano Banana Pro 1K or 2K$1.125$2.25
Nano Banana Pro 4K$2.25$4.50

Which one fits a label with fine print?

Ideogram launched 4.5 on 2026-09-30 and describes it as "the most precise edit model" that "eliminates artifact buildup" (Ideogram on X, read 2026-10-05). That is a statement about edits, not about packaging, and a back-of-pack panel of ingredients is a harder test than a front headline. Nano Banana Pro's 4K tier gives you more pixels for small type, and Ideogram's top tier on Sume is 2K.

So the split most buyers end up with is by text size. Front-of-pack headlines and flavor names go to Ideogram at low or medium. A dense back panel goes to whichever model reads clean at your print size, tested at that size, with the 4K Nano Banana Pro run as the control.

The ratio and shape rules

An Ideogram edit with no aspect_ratio returns the source geometry, according to the Image API docs. For Nano Banana Pro, send an aspect_ratio that matches your source (or auto, which is in its list) and confirm the result dimensions on the first file before you run the other languages. A flat dieline at 1:3 or 3:1 is not in the Nano Banana Pro list, but it is in Ideogram's.

Send PNG files, not JPEG, between passes. Re-compressing a label adds blur that a later pass can mistake for design.

A decision rule

Run one language on both, at the size you print. If Ideogram at low passes your proofread, you spend $0.0375 per language and stop. If it fails on small text, raise it to medium ($0.075) before you pay $0.1875 for Nano Banana Pro. If both fail, you have a design problem, and the fix is bigger type on the master, not a bigger model.

A small test before the full run

Before you scale this up, run it on two or three real files first and write down what you saw. A small test at low quality costs cents on Sume ($0.0375 per Ideogram 4.5 edit), and it tells you whether your prompt, your source files and your review step are ready.

Keep the originals untouched, name every output after its source and its prompt, and store the job id with each result. If a result is wrong later, you can find the exact request, fix the prompt and re-run only that item with a new Idempotency-Key.

Sources

Related posts

More in Comparisons

All Comparisons posts

Written by Sume