Image Captioner (MacOS)

A focused GUI to caption dataset images for LoRA/SDXL training. It streamlines annotation, guarantees consistent file naming, and exports images and captions in the exact format most LoRA pipelines expect: sequentially numbered images plus same‑named .txt files with prompts. Tested only on macOS Sequoia 15.5

Why this speeds up LoRA training

• Zero manual file management: the app renames images to 1..N and writes 1..N.txt automatically. No more hand‑renaming or misaligned captions.
• Fast per‑image editing: thumbnail + caption field per image, with a clear Done/Not done status. You always know progress.
• Batch operations: drag to select a group, then bulk Delete or Mark as done to clean or finalize quickly.
• Filter/sort and export only what you need: focus on items without captions or not done; export only the visible subset to iterate in small batches.
• Fewer mistakes: the tool guarantees matching names for images and captions, consistent encoding (UTF‑8), and preserves original extensions.
• LoRA‑friendly conventions: numbering and sidecar .txt files align with common SDXL LoRA scripts, so training can start right away.

Features

• Load an unlimited number of images
• Per‑image caption input with status: Done / Not done
• Delete per image + bulk delete or bulk mark done (drag‑select)
• Lightbox preview (click thumbnail; close on click outside or ESC)
• Filters and sorting by caption presence and status
• “Export visible only” respects current filters/sorting
• Choose Start Index; numbering is continuous (1..N)
• Language switch: EN/RU
• Minimalist, modern dark UI

Supported image formats: png, jpg, jpeg, webp, bmp, tiff, tif, heic, heif.

Installation (macOS)

Note

Node.js LTS (18+ recommended). Install from https://nodejs.org.

✦ Electron-app:

1. Download or clone this repository to any folder on your Mac, for example ~/image-caption.
2. To run Electron app open Terminal and run in target repository:

git clone https://github.com/DualChimerra/ImageCaptioner.git
cd ~/ImageCaptioner
npm install
npm run start

✦ Building DMG installer:

The installer will be generated in the dist/ folder (electron‑builder).

npm run dist

Usage

1. Click Load images and choose your dataset images.
2. Click Choose folder and set the output directory.
3. Optionally set Start index (default 1).
4. For each image, write a caption prompt; toggle status Done/Not done.
5. Select multiple items by dragging a selection rectangle (Cmd/Ctrl+click toggles individual items).
6. Use the bottom bar to Bulk Delete or Mark as done.
7. Use filters/sorting to focus work (e.g., “without caption”).
8. If you need to iterate, enable Export visible only to export just the current view.
9. Click Export. A modal will confirm how many files were exported and let you open the folder.

Roadmap ideas

• Windows support
• Drag & drop import
• CSV/JSON import/export of captions
• Autosave session and restore on launch
• Tag presets and autocomplete for common tokens

License

Custom Non-Commercial License (CNCL)