GPT Image 2 Creative Workspace
Upload a source image, keep the subject visible, and generate campaign-ready visual directions with GPT Image 2 image-to-image reference workflows.
GPT Image 2: 30 credits at 1K, 50 credits at 2K, 88 credits at 4K. Each output is charged separately. Failed outputs are refunded.
First image is the main reference. Tap any tile to promote it.
Every output is an independent render and is charged separately.

Matched prompt
I want to see what you really look like. Draw a snapshot of your everyday life as if it were accidentally taken on an iPhone. Make it feel like a very ordinary, imperfect candid shot. The photo should have slight motion blur, with uneven, natural lighting.
Conversion path
The workbench above is prefilled with a GPT Image 2 prompt for this use case. Generate a first direction, then reuse strong results or your own product and venue photos as references.
Yes. Uploaded images are stored securely for the generation request and passed to the selected reference-capable model.
Large supported files are compressed before upload when possible, then uploaded to storage before generation.