GPT Image 2, OpenAI's image model, is available in Photoshoot.si. It runs on the same credit balance as every other model in the app, so you can put it side by side with Nano Banana, Seedream or FLUX.2 on the same photos and look, and keep the one that works.

What it is

GPT Image 2 is built around precise instruction following and clean text inside the image. It suits prompts where the details matter: a look with several named garments and props, a product shot where the label has to stay readable, or an edit that changes one thing and keeps the rest.

What it is good for:

  • Long, precise instructions. Write the prompt as full sentences. The model follows them closely, so say what you want where.
  • Text that has to survive. Packaging, labels and signs in a product photoshoot.
  • Edits with references. Upload up to 6 reference images in the studio and describe how to combine them.

Where it runs in the app

  • Photoshoot and headshot tools such as AI Photoshoot, AI Headshot Generator and LinkedIn Headshot list it in the model menu, next to the default.
  • Product tools such as AI Product Photos list it too, for shots where the label text matters.
  • The studio lists every image model, GPT Image 2 included.

The sample above was made with GPT Image 2 from a text prompt.

Settings

GPT Image 2 makes images in 1:1, 4:3, 3:4, 16:9 or 9:16, and up to 4 images per run. The price is per image, so 4 images cost four times one, and a Full shoot costs one image per look; the button shows the total before anything is spent. It accepts up to 6 reference images; a tool that takes fewer photos keeps its own limit.

A cheaper sibling, GPT Image 1 Mini, is in the studio for quick drafts; move to GPT Image 2 when the text or the details have to be right.

Known limits

Failed generations are refunded automatically. Requests rejected by OpenAI's safety filter are refunded up to a daily limit, and prompts that our own check blocks never cost credits. The content policy has the details.

Try it

Open the AI Product Photo Generator, upload a phone photo of a product with a label, switch the model to GPT Image 2 and pick a look.