When working on long-term, multi-image visual projects (such as designing a 78-card Tarot deck, a comic book, or game assets), ChatGPT retains the text memory and decisions surrounding previously generated images, but it loses the ability to retrieve, reference, or redisplay the actual image files from earlier in the chat history or in separate branches.Currently, users must manually download every generation, store them locally, and constantly re-upload them to the chat just to maintain visual consistency or request revisions. This breaks the workflow and wastes user and system tokens.
I would like ChatGPT to have persistent access to a chat’s complete generated-image history. Specifically:Index & Retrieve: The ability to ask ChatGPT to “Retrieve the image generated for the ‘The Fool’ card from yesterday” and have it pull up that exact historical image asset.Side-by-Side Comparison: The ability to request a comparison between two previously generated images in the same thread (e.g., “Compare Card 1 and Card 5 to ensure the art style matches”).Continuous Revision: Allowing ChatGPT to natively reference the exact GenID or seed of a image generated dozens of turns ago to make consistent tweaks, without requiring a manual re-upload.
The current workaround requires managing an external file folder, tracking our own GenIDs manually, and uploading files back into the chat interface. This is highly inefficient for large-scale creative projects.
This feature would bridge the gap between ChatGPT’s excellent text-based long-context memory and its current image-handling limitations, unlocking massive value for creators, designers, and visual storytellers.
Thanks for taking time to consider this!
-A