Mexes

Notice: The Aibooru API has been quite unstable in recent days. I recommend using the Danbooru API provider.LINK: WEBV5: A new API provider has been added.Just that, a new API provider.Minor changes:Remove the option to choose between “Most popular” and “Most recent”; now “Most recent” will always be used.Replace the rating selector with a simple button that switches between safe and unsafe.V4: new option and minor fixes.I added the possibility of a new input called "Tags to add". Unlike "Tags to remove", this one adds the tags you want at the beginning of the prompt. It’s useful if you want to test a LoRa with a keyword, if you simply want to add your preferred quality tags, or if you want to add new elements to all prompts.Minor changes:Added some more metatagsMade it so that disabling the "Characters" option also removes tags that refer to a character (alternate costume, official costume, alternate hairstyle, etc.)Fixed an issue that left behind residual commas when removing quality tagsV3: More options!!!I’ve added the ability to use Aibooru.Why Aibooru?Most of the images people upload to Aibooru come with all their generation metadata, including the prompt.So, if you use the prompts from Aibooru, it’s much more likely that you’ll get results closer to the reference image.Why didn’t I use this from the start?Personally, I don’t like how some people prompt (they use tags that don’t exist, write messy prompts, or simply use natural language). That’s why I went with Danbooru at first.But hey, now this is a tool for the community and not just for me, and the more options the better, right?Options:Remove LoRa Tags: Sometimes prompts come with LoRAs included. This simply removes them if you’re not interested.Remove Quality Tags: This removes the quality tags that are used, like “score_9_up, masterpiece, super ultra mega quality, etc.”Why can’t I use Danbooru options on Aibooru?The way prompts are obtained on Aibooru is different from Danbooru. On Danbooru, the tags come from the site itself, where they’re used to filter searches—making it easier to know which tag refers to a character, copyright, etc. On the other hand, Aibooru pulls the raw prompt directly from the metadata, so the tags don’t carry any additional information.V2: I added many more options to better configure the prompt according to your preferences.Tags to remove: this option lets you enter any tag you want to remove from the final prompt. Don’t want generations with blue jackets? Just type “blue jacket” and that tag will be removed from the prompt. (this does not filter the results you get from the API; you will still receive images that contain blue jackets)Character: this lets you remove or add character tags to the prompt.Copyright: this lets you remove or add tags for the name of the franchise of the respective character/setting.Combine tags: this combines tags to reduce the size of the prompt. For example: if the prompt contains “skirt, white skirt, pleated skirt” this will combine everything into a single tag “white pleated skirt.”As for minor changes, I can say that:I changed the page colors so they wouldn’t burn your eyes.I added more tags to remove (some of them: web address, original, patreon logo).V1:I’ve been creating LoRA models and checkpoints, and I was spending too much time crafting prompts to test them thoroughly. That’s why I decided to develop this web app. At first, it was just a personal tool, but I soon realized it could be useful to many more people, so I refined it to make it easy and comfortable for anyone to use.What does this tool do?It generates prompts based on the tags from Danbooru posts.It automatically removes metadata and tags that restrict generation (for example, “white background” or “censor”).It filters out artist tags and other irrelevant ones, so you can copy and paste the prompt directly.LimitationsIt won’t always recreate the original image accurately, since it depends on the quality of the tags in each post.Occasionally, unwanted tags may appear. If you notice any, feel free to let me know!Found a bug?Don’t hesitate to tell me — I’m happy to fix them all.Have feature suggestions?Absolutely! Your ideas are welcome.Is the site down?The app runs on a free hosting plan, so sometimes resources get exhausted. I apologize in advance if that happens.Why a web app and not a desktop app?I noticed many users access from mobile devices, and I wanted it to be accessible to everyone, regardless of platform.Want to support me?Thank you so much! you can help me:Leaving a comment and rating.Using my modelsThat’s it!Stay hydrated and don’t forget to blink.

Translated using ChatGPTI spent several days exploring how to expand examples in datasets with few images, and I came up with the idea of using chaiNNer to automate the process. After testing different training configurations, I decided to share it so others can save time doing the same thing.What does this tool do?- Rotates the image by 15° and 345° (i.e., both directions).- Automatically crops the black borders that appear after rotation, using simple math.- Optional: it can also do a horizontal flip and save those images.That's all it does—but it’s useful. I used to do this manually and didn’t know of any other tool that automated it.How to install chaiNNer?Super easy:1. Download the installer from: chaiNNer2. Run the .exe3. Click “Next” on everything… and done!> ⚠️ You don’t need to download additional dependencies—those are only necessary for AI features like upscale, inpaint, etc.Download workflow: Workflowunzip and inside you will find the filechaiNNer stuck on infinite loading?Yeah, it happens to me too. What I do is install both the installer version and the portable one, and open both at the same time. For some reason, that makes the app load. Once it’s open, you can close either one with no issues.Do you know a better way to do this in chaiNNer?Let me know! I’m open to improving and updating this tool.Do you know other data augmentation techniques?I’d love to hear about them too. My goal is to build a more complete data augmentation tool (within chaiNNer’s capabilities, of course).That’s it!Stay hydrated and don’t forget to blink.