Last year, when I started messing around with NAI, I got a bit frustrated because managing tags was a mess, and adding or removing characters and actions could turn the prompt into a disaster real quick if you weren’t careful. So, I built a small javascript app to create and format the prompts the way I wanted
>The idea was to organize tags into separate groups, rearrange them manually using drag and drop, disable them when needed, or turn them into negative tags with just a click>I also added an option to export groups (or all of them at once) as image files, so I could quickly reload specific templates whenever I needed them (characters, outfits...)After seeing how v4 gave (kinda) full control over characters, I had a few ideas to improve the app, so I went all in:
>Now I can separate general tags (art style, scene composition...) from character specific ones (the character's gender, name...)>This allowed me to add contextual toggles, so I can assign quick tags to specific characters (expressions, hairstyles, outfits...), or even set the overall scene composition in just a few clicks (the view angle, the location, the mood, what the characters are doing...)>The app also cleans up the prompt by removing duplicate tags outside character groups and handles other small fixes so everything can work correctly>Finally, I built a browser extension that links my prompt builder to the NAI web app, letting me send the prompts directly and generate images without having to copy and paste the final promptNow I can set up any scene in under a minute and tweak whatever I want without having to edit the whole prompt. For example, the result in pic rel is...
>https://files.catbox.moe/ncvz91.pngAnd if I need to change something, I just toggle the relevant tags or adjust the groups...
>https://files.catbox.moe/s74g6b.png>https://files.catbox.moe/88x38z.pngv4 still does its own thing sometimes and occasionally misplaces character specific tags, but most of the time, it works pretty great. And yeah, this was all made for gooning
