GPT-5.6 Responses API, partial image streaming, masking & multi-turn edit context.
Multi-image blending (3 source slots), 2K resolution, aspect ratio overrides & style transfer.
Upload a primary image and, when the selected model supports it, as many as two additional references. The first image is the editable canvas. Supporting images can supply identity, clothing, texture, composition, product, or style references. Paint a tight mask on the primary image when using a model with hard-mask support.
Select one or more images, edit each description, then submit the batch to ImageKit and the gallery.
Descriptions are saved as the gallery labels for each uploaded image.