image

Smart Image Sorter

Sort up to 500 images with local AI, review uncertain results, and save the original files into category ZIPs.

Local processing

About your filesFiles and content selected in this tool are processed in your browser and are not uploaded to a Browser Kitty backend.

Operation demo See the basic workflow in a short video.

About this tool

Smart Image Sorter is a browser tool for sorting large sets of photos and images into predefined categories. Load up to 500 images, keep only the categories relevant to the current batch, and classify them with AI that runs on the device. The release UI uses a fixed category catalog rather than arbitrary free-text labels.

Uncertain AI results are separated into a needs-review queue instead of being mixed with confident results. You can correct one image or batch-change several, then save the original files without recompression into category ZIPs. Classification, review state, and ZIP generation stay on the device; selected images are not uploaded by the app.

Good for

Roughly separate hundreds of travel, event, or everyday photos into broad categories before manual organization
Reduce review work by letting AI handle clear cases while a person checks only uncertain images
Organize product, document, scenery, and other image types into ZIPs without converting the original files
Continue sorting a batch even when some images are damaged or cannot be decoded

What it can do

Key capabilities available in this tool.

Add up to 500 JPEG, PNG, WebP, or AVIF images individually or by folder
Choose only the fixed categories you need and sort images with local in-browser AI
Send low-confidence or close-call results to a review queue so you can focus on uncertain images
Change one result at a time or batch-change selected images with a confirmation step
Separate failed images from normal results with individual/batch retry and exclusion controls
Save original files without recompression into category ZIPs while avoiding duplicate names safely

How to use

Add JPEG, PNG, WebP, or AVIF images, or select a folder containing images.
Choose at least two fixed categories relevant to this batch; recommended categories are enabled initially.
Start sorting. The local AI compares each image against the selected category candidates and assigns a result.
If there are uncertain items, open the needs-review flow and move through them to confirm or correct the category.
If needed, also change normal results one by one or batch-change selected images.
When review is complete, save category ZIPs. The original images are included without recompression.

Supported

Input: JPEG, PNG, WebP, and AVIF through file selection or folders, up to 500 images per batch.
Classification: choose at least two predefined categories. The displayed match score is relative among the currently selected candidates.
Review: needs-review flow, individual changes, confirmed batch changes, and retry/exclusion for failed images.
Export: category ZIPs containing original files without recompression, with safe duplicate-name handling; one ZIP operation is limited to 750 MB of source images.
Work recovery: classification and manual corrections can be stored locally, but image files themselves are not saved, so the same images must be selected again.

Limitations and notes

A batch is limited to 500 images, and large numbers of high-resolution files can use significant device memory.
Available categories are fixed at build time; the release UI does not add arbitrary free-text categories.
Match scores are relative among the selected candidates, not probabilities of correctness. Review important sorting results manually.
Image decoding support varies by browser and OS, and failed images are not included in category ZIPs.
A single category-ZIP generation operation is limited to 750 MB of source images.
Work recovery does not save the image files or image embeddings, so the same images or folder must be selected again.

Frequently asked questions

Are images uploaded to a server for AI processing?

No. The distributed HTML contains the AI model and runtime needed for classification, review, and ZIP generation on the device. The hosted version only needs the initial page download.

Can I type arbitrary category names?

The release UI uses predefined categories. Choose the categories relevant to the current batch; adding categories requires updating the build-time category definition and embeddings.

Does a high match score guarantee the category is correct?

No. The score is relative among the currently selected category candidates, not a probability of correctness. Review important images manually.

Are images recompressed after sorting?

No. Category ZIPs contain the original image bytes rather than recompressed copies. Duplicate names are adjusted safely within each category.

Offline version and source code

The standalone HTML embeds the AI model and runtime, so a saved local copy can sort images offline without sending them out of the device.

View source on GitHub