Dagger is an application that strongly supports the creation of comma-separated caption files commonly used in Stable Diffusion. You can use it through your web browser.
It is designed to be simple yet powerful, enabling you to organize your dataset before LoRA training or FineTuning more efficiently.
Dagger treats an image file and its caption file with the same name as a pair.
Dagger is compatible with the output files from WD14-Tagger. You can pre-tag your data if necessary.
https://github.com/toriato/stable-diffusion-webui-wd14-tagger
You can add files via drag and drop or by selecting the files you want to import from the folder icon in the upper left corner. Dagger will search for a text file with the same name as each selected image file (.txt, .caption) and treat them as a pair. Separate inputs are not allowed.
Dagger provides several features to make editing convenient.
When you select an image, its details and tags will be displayed in the right pane. You can add or remove tags from here.
You can select multiple images using CTRL or SHIFT mode. When multiple images are selected, the right pane displays the common tags of all selected images.
You can use CTRL+A to select all images currently being filtered.
After importing images, all tags will be displayed in the FILTER BY TAGS at the bottom of the screen. You can filter images by selecting these tags. The number displayed next to each tag indicates how many images have that tag.
By clicking in normal mode, only images that have the selected tag(s) will be displayed.
When clicking in CTRL mode, only images that do not have the selected tag(s) will be displayed. If you click the delete button of a tag while in CTRL mode, that tag will be removed from all images.
You can trim images. Pressing Save will overwrite the image with its trimmed version. Pressing Save as New will save a copy of the image.
When you finish editing, save the edited images and caption files from the save icon in the upper left corner. If you click the save icon while having images selected, only those images will be saved. You can use Ctrl+A while applying a specific tag filter to output only the images subject to that filter. This feature is useful when you want to experiment with various learning variations.
Please use GitHub's Issue tracker for feature requests and bug reports. If you're in Japan, feel free to use Japanese.
Pull requests are also welcome. However, Dagger values simplicity. There's a chance we may not merge feature addition PRs, so please be aware of that. It's smoother if you issue beforehand.
MIT