Pinterest Performance+ Priority Products: build a creative set

Pinterest added Priority Products to Performance+ so advertisers can steer automation toward chosen products in moments like a seasonal launch, promotion or category push. The creative job that follows is small: a 2:3 image set for those few products, made once, instead of fresh art for every item in the catalog.

The feature description is from Pinterest's Pinterest Presents announcement, read 2026-10-01. It does not state image sizes; those come from Pinterest's ad specs. Sume parameters are from the Image API docs.

What did Pinterest announce?

The newsroom post says Performance+ is adding Priority Products, giving advertisers more control over the products they want to prioritize while keeping the scale and efficiency of automation. The same post announces self-serve A/B testing and Visual Search Ads.

How big should the creative set be?

Keep it proportional to the priority list. Say you mark 8 products: plan 3 images each, 24 images, instead of rendering for hundreds of SKUs. Sume's n parameter returns several images per call (the docs allow up to 10, with lower per-model ceilings; the catalog example shows 1 to 4), so one request per product with n: 3 gives three takes of that product's scene.

Which ratio and size do I ask for?

2:3 is in the Image API's normalized ratio list. For exact pixels on a GPT model, use image_size: both edges multiples of 16 and at least 655,360 pixels, so 1024x1536 works at 1,572,864 pixels. Confirm the final Pin size on Pinterest's specs, and see the 2:3 Pin image post for what is documented there.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: pin-priority-sku-101" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "Holiday gift tin on a wooden table, soft window light",
    "image_size": "1024x1536",
    "n": 3
  }'

What if the priority list changes?

Add the new product and render only its set. If you do need many at once, Format bulk runs queue up to 100 runs with a concurrency window. Each request keeps its own Idempotency-Key, so a repeat for the same product returns the original job. Variation across a creative asset is covered in asset-level variations.