Basic workflow
Open the Generator, enter the prompt, choose a quality mode, select an aspect ratio such as 1:1, 16:9, or 9:16, generate, review, upscale when needed, and download.
How to use GPT Image 2
Good GPT Image 2 results usually come from clear structure rather than long prompts. Separate the subject, exact text, style, lighting, composition, aspect ratio, and output use.
If the image includes visible words, paste the exact wording in quotes and keep it short. Treat the first generation as a draft, then iterate on layout and review details before publishing.
Key takeaways
Prompt structure
Write the subject first, then exact image text, visual style, composition, lighting, color palette, camera or design treatment, and any constraints to avoid.
Quality control
Before using an output, check spelling, numbers, UI labels, hands, faces, brand references, rights-sensitive elements, and whether the final size matches the channel.
FAQ
Should prompts be written in English?
English works well, but you can prompt in other languages. If the image must contain non-English text, paste the exact characters and state the target language explicitly.
How much text can I put inside an image?
Short headlines, labels, and prices work best. Long paragraphs and dense legal text should be added manually after generation.
When should I use Ultra quality?
Use higher quality for client work, print, packaging, posters, or assets you plan to upscale. Use draft quality for early composition exploration.
Can I reuse gallery prompts?
Yes. Use them as structure, then replace the subject, wording, brand-safe details, style, and aspect ratio for your own use case.
Start with a proven prompt structure
Pick a gallery example, copy its structure, then replace the text and output format for your campaign or product.
View gallery