52 weekly podcast covers: $5.20 on Nano Banana 2.1, $1.29 on GPT

A year of weekly podcast covers is 52 square images: $5.20 on Nano Banana 2.1 at 1K or $1.29 on GPT Image 2.5 low on Sume, with the arithmetic shown.

4 min readSume
All posts

The short answer

A year of weekly podcast cover art is 52 square images. On Sume that costs $5.20 with Nano Banana 2.1 at 1K (52 x $0.10) and $1.29 with GPT Image 2.5 low at 1K (52 x $0.02475 = $1.287). At 0.5K, Nano Banana 2.1 is $3.90.

The cost of the images is small next to the time spent choosing them, so the useful question is how many drafts per episode you can afford.

The arithmetic

Directories ask for large square covers, and a 1K square is the natural size to generate. Adding two extra takes per episode multiplies the totals by three: $15.60 on Nano Banana 2.1 and $3.86 on GPT Image 2.5 low.

The table prices one image per episode. The Google row shows what Google's own page lists for the same volume of 1K images.

52 episode covers, as of 2026-10-08 (totals rounded to the cent; the Google row is a reference and not sold through Sume)
OptionCount x unit priceTotal
Nano Banana 2.1 at 1K52 x $0.10$5.20
Nano Banana 2.1 at 0.5K52 x $0.075$3.90
GPT Image 2.5 medium (2K)52 x $0.055625$2.89
GPT Image 2.5 low (1K)52 x $0.02475$1.29
Google direct, 1K standard (not on Sume)52 x $0.0336$1.75

How to run it

Keep the series look by sending last week's approved cover as an input reference with the new episode title in the prompt. Check the input_references range for the model you chose before you rely on more than one.

Set a fixed model id in the script. If you use sume/auto, the model that ran is not disclosed, and a series needs a stable look.

Retries and cost per episode

Some images will be rejected by eye, and rejected images are billed like any other, so budget a margin. The bullets add 20 percent to each Sume total in the table above (the totals multiplied by 1.2) and divide by the number of episodes to give a per-episode cost.

Treat the per-episode figure as the number to compare across models, because it does not change when the count changes. Re-read the public catalog before a large run, since prices can change and the table is a snapshot from 2026-10-08.

  • Nano Banana 2.1 at 1K: $5.20 becomes $6.24 with 20 percent extra images, which is $0.12 per episode (a 52-episode job).
  • Nano Banana 2.1 at 0.5K: $3.90 becomes $4.68 with 20 percent extra images, which is $0.09 per episode (a 52-episode job).
  • GPT Image 2.5 medium (2K): $2.89 becomes $3.47 with 20 percent extra images, which is $0.07 per episode (a 52-episode job).
  • GPT Image 2.5 low (1K): $1.29 becomes $1.55 with 20 percent extra images, which is $0.03 per episode (a 52-episode job).

What the docs say to check

Passing model sume/auto lets Image Router choose the family and Sume never discloses which one ran, so a budget built on a fixed price per image needs an explicit model id.

Nano Banana 2 is retired on Sume. The ids google/nano-banana-2 and nano-banana-2 still work and run as Nano Banana 2.1, and the job stores the 2.1 id, so old code keeps working and the 2.1 prices apply.

Archive the returned URLs somewhere you control. The Sume URLs are signed, so store the file itself when it is part of a published feed.

The numbers above are list prices read on 2026-10-08, not measurements. Sume figures are the catalog prices that already include its margin, Google figures are from its own pricing page, and quality or speed differences are not covered. Re-read both pages before you commit a large budget, because either side can change its rates.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume