Studio Image Generation

How to create images in Studio with Gemini and OpenAI GPT Image 2.

Overview

Studio lets you create campaign images, product visuals, social assets, and website creative from a prompt. You can start from text, connect a reference image, choose an aspect ratio, and select the image model that best fits the job.

Studio can also generate a curated campaign pack from a selected image. See Studio Campaign Packs for the full workflow.

Available Models

Gemini Flash

Use Gemini Flash for quick image ideas and fast drafts.

Gemini Pro

Use Gemini Pro when you want higher-quality image generation and larger outputs.

GPT Image 2

GPT Image 2 is OpenAI's image model, also known from OpenAI's ChatGPT Images 2.0 release. In Studio it appears as GPT 2 in the model selector.

Use GPT Image 2 when you want strong prompt-following, polished campaign images, and reference-led creative work. It can use a connected image node as a visual reference, which is useful for variations, new layouts, and brand-safe creative directions.

How To Use It

  1. Open Studio.
  2. Add a prompt node or select an image node.
  3. Open the image settings.
  4. Choose the aspect ratio.
  5. Choose the model.
  6. Choose 1K, 2K, or 4K quality.
  7. Generate the image.

If you connect an existing image into a prompt node, Studio sends it as a reference image. The new image is generated from both the reference and your written prompt.

Quality Settings

  • 1K is best for fast drafts and lower cost.
  • 2K is the default for polished web, social, and campaign-pack drafts.
  • 4K is best for final campaign images or selected high-quality reruns.

GPT Image 2 maps these to OpenAI's low, medium, and high quality settings behind the scenes.

Campaign Packs

When an image node contains a product or reference image, Studio can open a campaign-pack composer. You can choose the editorial direction, shoot location, and which shots to include.

The default pack creates seven assets: website hero, product detail lifestyle, social square, story or reel, email banner, paid ad, and editorial detail. Studio uses GPT Image 2, the selected image as the reference, and the existing creative cost controls.

The default 7-image 2K pack is estimated at about $0.29-$0.37 in OpenAI output cost, plus text and reference-image input tokens. Selected winners can be rerun at 4K for final review.

What To Expect

GPT Image 2 can produce strong visual concepts, but generated images still need human review before publishing. Check text, logos, product details, layout, and any claims in the final image.

Some prompts may be blocked for safety reasons. If that happens, rewrite the prompt in a clearer and safer way, then try again.

Setup

GPT Image 2 requires an OpenAI API key in the server environment:

OPENAI_API_KEY=...

Do not paste API keys into prompts, pages, docs, or source code. Studio only reads the key from the server environment.