GPT Image 2: understands complex requests
Models1 min read
OpenAI's image model. Its strength is reading a long, tangled prompt in full without losing details: "the boy on the left holds a red balloon, a dog stands on the right, a bridge in the background".
Open in Studio: GPT Image 2 · GPT Image 2 Edit
- The model-name button — click to open the list.
- Every version of the family with its price; the tick marks the current one.
| Version | Price | What it does |
|---|---|---|
| GPT Image 2 | 10 cr. | Image from text |
| GPT Image 2 Edit | 10 cr. | Text-driven editing, up to 16 source images |
When to pick it
- The scene has many objects and relations between them.
- You need accurate text in the frame and clean typography.
- You need to merge several photos into one frame — the Edit version takes up to 16 at once.
A nice detail
Resolution 1K, 2K and 4K all cost the same — 10 credits. You can go straight to 4K without paying extra.
Tip
Write the prompt like a script: list the objects in order and say where each one is. GPT Image 2 follows that kind of instruction better than any other model in the catalogue.