Selfie or full-body photo for AI try-on: what to send

What the launch coverage says

TechCrunch, read 2026-10-03, reports that ChatGPT's Try On lets people upload a selfie or full-body photo to see clothing and accessories on themselves, that a Try On button appears in shopping results, and that people can upload an item image such as a screenshot and ask for a try-on. It says the feature uses ChatGPT Images 2.5 and that photos are saved in a Library alongside the try-on images.

The article gives no guidance on which photo works better, so treat the rest as general practice rather than an OpenAI rule.

Which photo suits which item

The same idea through Sume's image API

Sume's Image API lists ChatGPT Image 2.5 as openai/gpt-image-2.5 with up to 16 image references. Send the person photo first, the garment second, and say what to keep. Both must be public HTTPS URLs; localhost, private and non-HTTPS URLs are rejected, per the Media inputs rules.

Use aspect_ratio: "auto" on an edit so the output matches the person photo; the docs say omitting the field is not the same as auto.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: tryon-001" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "Dress the person in image 1 in the jacket from image 2. Keep face, pose and background unchanged.",
    "input_references": [
      {"type": "image_url", "image_url": {"url": "https://example.com/person-full-body.jpg"}},
      {"type": "image_url", "image_url": {"url": "https://example.com/jacket.jpg"}}
    ],
    "aspect_ratio": "auto"
  }'

Consent and honesty

Use photos of people who agreed. A try-on image shows a garment on a body, not a promise of fit, so keep the size chart next to it.

Check the result against the garment

Compare the output with the garment photo for colour, pattern placement and sleeve length before using it. Try-on images are generated, so details such as a logo or a print can drift. If something is wrong, tighten the prompt and say what to keep, rather than adding more references.

Keep the person photo and the generated result together in your own records, with the prompt, so you can show how an image was made if a platform or customer asks.