8 YouTube chapter cards: GPT Image 2.5 low is $0.004 each at 1280x720
Eight 16:9 chapter cards at 1280x720 cost $0.032 on GPT Image 2.5 low via Sume, against $0.20 to $0.80 on flat-price models. The request and the math.

Eight 16:9 chapter cards at 1280x720 on ChatGPT Image 2.5 with quality: low bill about $0.004 each on Sume, $0.032 for the set before cent rounding. That is among the cheapest 16:9 options in the catalog for this job, because GPT Image 2.5 is token-billed and a 1280x720 frame is a small output. The Image API docs explain the pricing basis: output image tokens at $30 per million, plus Sume's margin.
Chapter cards are a draft-quality job. They are title slides between sections of a long video, they sit on screen for a few seconds, and you will redo half of them. That makes the low tier the right default and the rest of this post the reason why.
How do I request a 1280x720 card?
Send image_size as explicit pixels. On GPT custom sizes both edges must be multiples of 16, the longest edge is at most 3840, and the image must hold 655,360 to 8,294,400 pixels. 1280 is 80 x 16 and 720 is 45 x 16, so it passes, with 921,600 pixels. One call, one card:
curl https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-image-2.5",
"prompt": "Chapter title card: bold white text 'Part 3: Pricing' on a deep blue gradient",
"image_size": {"width": 1280, "height": 720},
"quality": "low"
}'What do eight cards cost on each model?
GPT Image 2.5 is the only row that changes with size and quality. The others are flat per image. All eight cards are 16:9, which every catalog model lists.
| Model | Per card | Eight cards |
|---|---|---|
| GPT Image 2.5, low, 1280x720 | $0.0040 | $0.032 |
| Qwen Image | $0.0250 | $0.200 |
| Seedream 5 Lite | $0.04375 | $0.350 |
| Nano Banana 2 (1K tier) | $0.1000 | $0.800 |
When is low not enough?
Low quality is a draft tier. At 1280x720 the same card is about $0.00925 on medium and $0.0356 on high, so a medium pass over the eight cards you keep adds under 8 cents. Spell-check any text in the image before publishing, whatever the tier; image models misspell titles, and the fix is a regenerate or an overlay in your editor.
If you omit quality, Sume defaults GPT Image 2.5 to high, so a card you forgot to tag costs about nine times the low rate. Prices here are list x 1.25 in dollars per image before whole-cent rounding; the catalog's billable formula reads "list x 1.25, ceil usd cents", so confirm the first charge in usage.cost and budget from the invoice, not from the sum.
Sources
Related posts
More in Pricing
- Agent 365 cost management for Copilot Studio agents vs a per-run cap
Agent 365 sets spend policy per user group and adds Copilot Studio agents in October. A per-run cap on the API call bounds the one task. Know which you need.
- AI asset unit-cost sheet for client quotes: image, voice, video
A ten-row cost sheet from the Sume catalog: image, voice, transcript, cutout, music, render and video-second prices, all-in with the 5.5% fee.
- What do 50 AI image cards cost at low, medium and high quality?
Fifty Ideogram 4.5 images cost $1.875 at low, $3.75 at medium and $13.75 at high by Sume's catalog line. Read usage.cost on each result to confirm.
- AI video price per second, Oct 2026: Seedance, Wan, Kling, H3 on Sume
Billed price per second on Sume for Seedance 2.5, Wan 3.0, Kling 3 and MiniMax H3 at each resolution, with 5, 10 and 15 second clip totals.
Written by Sume