[Feature Request] Dataset-Wide Tag Index + Per-Image Tag Hiding

[Kitzune]

Feature Request: Dataset-Wide Tag Index and Per-Image Tag Hiding in Tag/Caption Editor

Summary

Add a dataset-wide tag index based on the currently selected input folder, and allow hiding tags on a per-image basis in the Tag/Caption Editor without removing them from the dataset or caption files.

Problem

In large datasets, the effective tag vocabulary grows significantly over time. Currently:

• There is no global view of all tags used across the input folder.
• Users must repeatedly re-evaluate tags that are known to be irrelevant for a specific image.
• This creates visual and cognitive clutter when editing .txt / .caption files.

This inefficiency becomes particularly pronounced when:

• Revisiting older images
• Expanding the dataset
• Working with auto-generated tags or captions

Proposed Features

1. Dataset-Wide Tag Index

Generate an index of all unique tags found in captions within the selected input folder.

Scope & Behavior:

• Source: Existing .txt / .caption files only
• Read-only overview (no automatic modification of captions)
• Optional additional data:
  • Tag frequency across the dataset
  • Awareness of exact-match vs. partial-match (consistent with existing filter behavior)

This serves purely as a dataset tag overview and does not replace per-image tags.

2. Per-Image Tag Hiding

Allow users to hide selected tags for the currently viewed image in the Tag/Caption Editor.

Key Properties:

• Hidden tags are not removed from the dataset or caption file
• Hiding is image-local only (does not affect other images)
• Hidden tags remain in the underlying caption file but are visually suppressed in the editor for that image

Example:

• Dataset contains tags: solo, 1girl, red dress, green dress
• Current image caption: solo, 1girl, red dress
• User can hide green dress for this image to avoid repeatedly evaluating its relevance

Benefits

• Significantly reduces tag clutter in the Tag/Caption Editor
• Speeds up manual review and correction passes
• Improves usability for large or long-lived datasets
• Works seamlessly alongside existing filtering and caption editing features

Optional Extensions (Non-Blocking)

• Toggle to temporarily show/hide hidden tags
• Visual indicator for hidden tags (e.g., faded opacity, icon, or color)
• Store hidden tags in a separate sidecar file or metadata (to persist across sessions)
• Batch action: hide a tag across multiple selected images

This feature would greatly improve workflow efficiency for users managing large datasets. Thank you for considering!

[Particle1904]

Hi, I've read the suggestion and I'll keep it in mind for future releases, I'm not actively developing the tools at the moment and only really implementing bug fixes for issues I find myself.
Tag frequency is already implemented and can be seen in the tag processing tab, only really works for exact matches, tho. I'll keep the issue open to keep track of the feature request.