Image Generator is a user-facing Tools application for creating and managing generated images. It is separate from OpenAI Engine administration.
/image-generator is the general entry point. Signed-in users are sent to their personal Image Generator workspace. Signed-out visitors can use guest generation only when guest generation has been explicitly enabled; otherwise they are sent to sign in./me/image-generator is the signed-in user's workspace for generation, owned image history, follow-up work, sources and publication controls. Existing backend ownership checks continue to scope ordinary users to their own records./me/image-generator/publishing is the owner's live publication manager for image, prompt, reference/source and category visibility./me/image-generator/profile contains personal Image Generator defaults, including the opt-in setting that can publish future Image Generator results immediately./admin/image-generator is reserved for global Image Generator archive and publication administration and requires administrator access./openai/images/archive remains the public image archive where public publication is enabled.Image Generator results are private by default. The personal profile setting Automatically publish new Image Generator images is also off by default.
When the owner enables it, future results stored for that user by the signed-in Image Generator workspace, archive follow-up generation and Image Generator API are created with the image, prompt and references/sources public. It does not retroactively publish existing archive entries, assign a category, publish anonymous guest images or change unrelated integration defaults such as Slack image generation. X-bot images keep their separate forced-public rules.
The publication manager saves ordinary row-level visibility and category changes over authenticated AJAX and reflects the authoritative stored state without reloading the page. Bulk publication uses the same live behavior when JavaScript is available and retains normal form submission as a compatibility fallback. Owner and administrator authorization is enforced by the backend for every mutation.
The signed-in workspace can keep several image generations active at the same time. Submitting another prompt does not cancel or replace jobs that have already been accepted.
Active jobs are shown separately with their own progress. A generated image is added to the completed-results area only after that specific job has completed successfully, so one slow or failed request does not block the others.
Accepted generation jobs run on the server-side job path rather than depending on the browser page remaining open. The workspace stores the accepted in-progress job references in the browser and resumes polling them when the same workspace is opened again. Terminal, expired or inaccessible jobs are removed from that local list. Backend ownership checks still apply to every status request.
Signed-in users can find Image Generator from /me and /services. Administrators can find the global archive and publication controls from /admin. The Image Generator guide is also listed in the public documentation catalogue.
The public archive is globally ordered with the newest published images first. User and category remain visible on each image card, and the archive can be filtered by user or category. Images are not split into contiguous user/category sections because that would conflict with global newest-first ordering.
The historical /admin/openai/images... web routes remain available during the compatibility period. Existing image-generation API and callback paths are not moved by this web-surface migration.
Anonymous generation is disabled by default. Making /image-generator available does not automatically make provider-funded image generation public.