OpenAI Image 2 (GPT Image 2) is OpenAI's latest image model. Compared to DALL·E 3 it follows complex instructions more accurately, renders text inside images more reliably and supports image editing with a reference photo (image-to-image). It is the same model that powers ChatGPT's native image features.

Yes. Click the image icon in the chat to attach up to 4 reference photos, then describe the edit you want — 'put this person in a beach scene', 'recolour the dress red', 'remove the background'. The model uses the OpenAI images.edit endpoint to keep your subject consistent while applying the change.

Be specific about subject, environment, lighting and style. GPT Image 2 is strong at following long, structured prompts — describe the camera ('shot on 35 mm film'), the mood ('moody, overcast') and the composition. After the first draft, iterate one change per message to keep the result coherent.

Yes — you own the output and can use it in ads, social posts, blog headers, merchandise, books and client work. We block obvious trademark and named-celebrity prompts so your output is safe to publish. Always review against your local IP rules for regulated industries.

Both tools share the same chat interface — switch models with one click. GPT Image 2 (OpenAI) is stronger at text rendering, instruction following and photoreal editing. Gemini (Nano Banana) is faster and cheaper for casual generation. Pick whichever model fits your job; pricing is the same per image on our platform.

Uploaded reference images are stored on encrypted volumes, processed over HTTPS, and auto-deleted within 24 hours. We never use customer images for model training. Generated outputs are kept tied to your conversation history so you can come back and download them later.