Snap DPA for Sponsored Snaps: catalog images to fix with Sume
Snap's Dynamic Product Ads now run in Chat for all global advertisers. The creative comes from your catalog, so the work is catalog images Sume can fix.

Snap launched Dynamic Product Ads (DPA) for Sponsored Snaps on September 10, 2026, and says they are available for all global advertisers (read 2026-10-03). DPA draws on your product catalog and conversion signals, so the creative you control is mostly the catalog's images.
Sume can make and fix those images through the Image API. It cannot upload to a Snap catalog or configure a DPA campaign.
What did Snap announce?
The Snap post is a launch announcement with some study numbers. The ones that matter here are the format and what DPA is built from; the performance figure is Snap's own and comes from its study.
- The 22% figure is Snap's study, not a promise for your account.
- The page gives no image size or file rules for DPA.
- Check Snap's catalog documentation for the feed fields before you upload.
| Item | Snap says |
|---|---|
| DPA for Sponsored Snaps | Dynamic Product Ads in Chat, launched September 10, 2026, for all global advertisers |
| Performance claim | 22% improvement in 7-day CPA when optimized for Pixel Purchase, in a July 17-26, 2026 study with autobid and 7-day attribution |
| Chat usage | 85% of Snapchatters use Chat regularly |
| How DPA picks products | AI recommendation models using advertiser catalogs and conversion signals |
| Other launches | Value Optimization beta, third-party web optimization open beta, omnichannel optimization alpha |
Which catalog images should I fix with Sume?
Because the model chooses which product to show, every SKU's image has to stand alone. Weak images are the common failure: shadows, clutter, a cropped product. The Image API takes up to 16 reference images on the GPT Image 2.5 models and an image_size with both edges multiples of 16.
A catalog clean-up job is one request per SKU with the original photo as a reference. The reference must be a public HTTPS URL, so host the file first.
curl -sS -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{"model":"openai/gpt-image-2.5",
"prompt":"Same product on a clean light background, soft shadow",
"input_references":["https://example.com/sku-1001.jpg"]}'How do I check what the model changed?
Check the output against the original before it goes into a feed. An image model can alter a label, a logo or a proportion, and a catalog image that shows something your product does not do is a policy and trust problem. Keep the original as the fallback for any SKU that fails review.
What Sume does not do
Sume does not read your catalog, post to Snapchat, or touch a DPA campaign. Snap's page does not state image specs either, so this post gives none. Generated images cannot return a transparent background, so ask for a plain white or light one instead.
Sources
Related posts
More in Use cases
- Swap multiple characters in a video with AI: 2 to 4 people
Swap several people in one clip with H3 Max Recast on Sume: one photo per person, left to right by default, up to four, and a price that ignores headcount.
- Template bulk edits on YouTube Shorts: what to vary per row
YouTube's Oct 1, 2026 originality update names template-based bulk changes as not original. How to make each row of a Sume bulk run differ in substance.
- Where GMV Max creative videos come from, and where AI clips go
TikTok Product GMV Max only runs videos from connected accounts, Spark Ads posts or ACA, with a product anchor link. Where a Sume-made MP4 has to go first.
- If TikTok removes an AI clip, are edited variants caught too?
TikTok says it will try to catch similar versions of misleading AI video that violates policy. What that scope is, and why bulk variants are a risk.
Written by Sume