Image-to-image from a reference avatar: six poses on GPT Image 2.5

To make a pose sheet from one reference avatar image, send the image URL in input_references and describe each pose in the prompt. Six stills on ChatGPT Image 2.5 at 1024x1024 with one reference cost about $0.125 at medium quality and $0.50 at high, on Sume. The Image API docs cover input_references, which takes up to 16 images on this model.

This is plain image-to-image on the image API. It is a different thing from a talking-head video: you get stills you can drop into thumbnails, covers and decks.

How do I keep the character the same?

Describe what must not change in words that are easy to check: hair colour, outfit, the logo on the shirt. Then vary one thing per prompt, the pose or the background. The model sees your reference but nothing guarantees identical output, so review all six stills before they ship. If you use a real person's likeness, make sure you have their permission.

{
  "model": "openai/gpt-image-2.5",
  "prompt": "Same character, same outfit and colours, waving with the right hand, plain pastel background",
  "input_references": [
    {"type": "image_url", "image_url": {"url": "https://example.com/avatar.png"}}
  ],
  "image_size": {"width": 1024, "height": 1024},
  "quality": "medium",
  "n": 3
}

What does a six-pose sheet cost?

Send two calls of n: 3, or one of n: 4 and one of n: 2. The price per still depends on quality and on the one reference's estimated input tokens.

Which tier should I use?

Draft at low, pick the two or three poses you like, and rerun those at high. The low estimate above is rounded from Sume's own calculator; read usage.cost on your first response to confirm. 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.