APP LINK: Booru Prompt Gallery - By Mexes
CHANGELOG: Web App: Booru Prompt Gallery Changelog | Civitai
TL;DR: Booru Prompt Gallery v9.0 adds six major features:
Import your own prompts to run them through the app's cleaning system
Smart Tag Exclusion to automatically avoid conflicting tags (e.g. face tags on from behind shots)
Variations/Wildcards mode to randomize tag groups between generations using
{ promptA | promptB }Random button for Merge and Variations modes to get random prompt combinations
Background Options to keep, remove, or replace background tags across all cards
Character Popularity Indicator showing how many Danbooru posts a character has, as a fidelity hint
Help Tooltips added throughout the UI for clarity
Hi everyone! I've added some interesting features to Booru Prompt Gallery, and here's the breakdown of what's new in version v9.0 of the web app.
Don't know what Booru Prompt Gallery is? I invite you to read the main article where the entire web app is explained in detail.
Web App: Booru Prompt Gallery - Get Prompts quickly! (V8.0) | Civitai
Before jumping into the new features, I'd like to mention that it's already been 9 months since I uploaded the first version of this web app, and it's great to see how much it has evolved over time. I'm also grateful to the people who use the site daily and to those who take the time to report bugs — they're the reason I keep maintaining and updating it. A big thank you as well to everyone who has contributed to tag categorization on the site.
This has been a project that has personally helped me streamline my AI content creation workflow and LoRA testing a lot, and it makes me happy to know that other people have found it useful too.
What to expect in the coming months?
I'm currently planning to build some kind of prompt creator, looking to take advantage of the entire ecosystem that has been built around the site. The idea is to go from being a place where you only grab prompts, to one where you grab them, create them, and share them.
Got an interesting feature in mind?
Say it! You can do so in the comments of this article or by using the feedback button on the site.
Want to support the site's development?
You can do so on KoFi. I also appreciate receiving any amount of buzz
Import and Clean Your Own Prompts
I've integrated a feature so you can clean your own prompts and extract image prompts to run them directly through the web app's cleaning and classification system. I built this to take advantage of everything the system has been building — not just with posts pulled from the API, but also externally with manually written prompts.
Smart Tag Exclusion
I kept running into the problem where tags added in "Tags to Add" would clash with compositions that had no relation to them. A clear example is from behind, where the prompt is based on the character's back but "Tags to Add" includes things that describe the face. To automate a fix for this, I implemented a small system that automatically disables those conflicting tags in prompts where they shouldn't appear.
This feature is still in beta, as I'm still refining the edge cases that can come up, but for now it works for face-related tags.
Variation Mode
Taking advantage of the interface and the system from merge mode and categorization, I implemented a variations/wildcards system. You can group several tag groups into a single prompt and one will be picked randomly during generation, using the format { promptA | promptB }. It's very useful for those who generate multiple images with the same prompt and want to vary things like poses between generations.
Note: I believe this tag format is only available for ComfyUI, so depending on your generation environment it may not work for you.
Currently you can use it by putting everything in Tags to Add and simply adding the category you want to randomize. It's a bit rough doing it this way, but I couldn't think of a cleaner way to integrate it without forcing you to use the card prompts. I'll try to improve this in the future!
Random Button for Merge Mode and Variations Mode
I've added a button to get random combinations and variations from the cards loaded during the session. If you want to use this feature, it's recommended that you scroll down first to load more posts.
You can configure it to keep certain categories fixed while randomizing the rest, as well as choose how many posts will be used to compose the final prompt.
Background Options
A constant problem I ran into was background generation. In my work environment, the LoRAs I use aren't great at that, so most of the time I prefer a completely solid background. To automate this I implemented this option with three modes:
Keep Original: Keeps the card unchanged in terms of background.
Remove: Removes all Scenario/Background tags from the prompt.
Replace: A text box appears where you can set your own background for all cards.
Character Popularity Indicator
Since Illustrious is trained on Danbooru, a great indicator of how accurately a character will come out without LoRAs is seeing how many posts that character has on Danbooru. Before, I had to go to the site manually to check — so I implemented a badge directly on the card that shows that information instantly.
If the prompt has more than one character, the indicator from the most popular one is used. This badge toggles on and off along with the Include Characters option.
Help Tooltips
The application has grown quite a bit in terms of options, so to avoid confusion I've added small info icons throughout the entire interface. Just hover over them and they'll quickly explain what each button or setting does.
And lots of Technical Changes and Bug Fixes
That’s it!
Stay hydrated and don’t forget to blink.
