ChatGPT Images templates (Poster, Merch): are they in the API?
ChatGPT Images 2.5 templates like Poster and Merch are a ChatGPT feature. Sume's image API has no template field, so write the brief as a prompt.
Not as a parameter. On OpenAI's launch page, templates such as "Poster" and "Merch" are a feature of ChatGPT: you pick a template and add the information, design elements or styles you want. Sume's Image API docs, read 2026-09-29, list no template field for POST /v1/images, so you recreate a template as a structured prompt.
What does OpenAI say templates do?
The page says templates make it easier to start creating across popular formats, like flyers and product photos, and describes choosing a template like "Poster" or "Merch" and adding details to personalize the result. It presents this as one of the product features it is introducing in ChatGPT, next to Sketch, comments on images and shared prompts.
| Feature | Where OpenAI describes it | Field in Sume's image docs |
|---|---|---|
| Templates (Poster, Merch) | In ChatGPT | None listed |
| Sketch | In ChatGPT, via @Sketch | None listed |
| Reference photos | Model capability | input_references, up to 16 |
| Focused edits | Model capability | Prompt plus optional mask_url |
How do I get the same result through the API?
Write the template as text. A template is a starting frame for a format, so put the format, the information to convey, the design elements and the style in one prompt, and reuse that text as your own template in code.
- Format: say "poster" or "merch mockup" at the start.
- Information to convey: put exact headline and date words in quotes.
- Design elements: colors, layout, and what must appear.
- Style: the look you want, or a reference image for it.
What does a poster request look like?
Store the template as a string with slots and fill it per request.
curl -X POST "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": "A poster. Headline: \"Night Market\". Date line: \"Oct 12\". Bold teal and orange, centered lantern illustration, lots of empty space at the bottom."
}'Can I keep a template's look across many images?
Send the same reference image and the same prompt skeleton on every call. The docs allow up to 16 references per request. Each call is a separate generation, so consistency depends on your inputs, not on a saved template. The same approach is used in GPT Image 2.5 reference images.
Will Sume add templates?
The docs do not say. Check GET /v1/images/models and the Image API page for the current parameters.
Should I keep templates in code or in the prompt?
Keep them in code as plain strings with named slots, such as headline, date and palette. That gives you the same repeatable start OpenAI describes for the ChatGPT feature, and each filled string is just the prompt you send. Review the output of each template once, then reuse it. Because completed generations are billed, test a template at a lower quality value before running a batch at the level you need.
Sources
Related posts
More in Use cases
- Custom AI avatar from a photo: allowlist or self-serve?
Google gates Live Avatar custom avatars behind enterprise allowlisting. Sume's avatar docs list a photo input on the create route. Request and rules.
- Customer service training videos: role-plays made with AI
Customer service training videos as short AI role-plays: which scenarios to script, how to build a two-person scene, captions, and what each clip costs.
- Dental marketing videos with AI: what to make, what to skip
Dental marketing videos AI can make without a patient: a meet-the-dentist clip, first-visit and FAQ answers, and office-photo tours. What to skip, and costs.
- Does an AI video API add a C2PA record or watermark to output?
Sume's docs do not say its video output carries a C2PA record or watermark, and trim and Timeline re-encode files. How to check a file and what to do instead.
Written by Sume