Easiest way I can think to do it would be to have an export to Google Photos or something of the ilk and then you put the prompt as the caption automatically. There's a much smarter way that I've been brainstorming with gpt4 that would involve a desktop app that would use NLP and image recognition to be able to make a really super relational database that is tagged not only by the descriptions and prompts but by AI analysis of the image so that things that were similar could be brought up or sorted by. Some kind of date tag which we already probably have would also allow people to follow their progress and to iterate back and forth pretty easily. They also recommended that you could put the prompt in the exif as an option, and point people to the tools that can use that to read and use them.