LorenzoDainelli 发布本地 AI 照片视频分类器 local-ai-frame-classifier 2.0.0-beta.143
LorenzoDainelli/local-ai-frame-classifier-rilasci
LorenzoDainelli 发布 Windows 本地 AI 照片视频分类工具 local-ai-frame-classifier 2.0.0-beta.143,安装包 38 MB,AI 模型全部在本机运行,照片、标签和姓名不上传服务器。
Sort your photos and videos with AI, entirely on your own computer.
⬇ Download for Windows · version 2.0.0-beta.143 · 38 MB · all versions · Italiano
Beta. The program works and updates itself, but it is still being tested. The interface is in Italian; these guides explain every screen in English.
Why
Most photo organisers upload your pictures to someone else's server before they can recognise anything in them. This one doesn't. The AI models run on your PC, your library stays on your disk, and nothing you label ever leaves your computer.
What it does
- Imports photos and videos (HEIC, JPEG, PNG, MOV, MP4 and more) and never touches the originals. It spots duplicates, pairs Live Photos with their video and reads the real date of every shot, time zone included.
- Learns your own categories. You build the tree of categories; the program studies the shots you have confirmed and suggests labels for the rest, always showing why: "it looks like these five you already labelled".
- Fast labelling from the keyboard, one shot after another, without the page ever reloading.
- Finds faces and groups the similar ones, so you name people one group at a time.
- Searches by words, with a language model that also runs locally.
- Tells you where you stand: how much is labelled, how often the suggestions are right on your photos, and what is worth labelling next.
- Optional encrypted vault: without the password, the library on disk is just noise.
- Dispositivi › Altro computer (Devices › Other computer): carry your work from one PC to the other on a USB stick. Nothing travels over the network.
- Enhances small photos and videos up to Full HD, locally, and the original stays untouched until you approve, comparing before and after side by side. The blind challenge lets you pick the better version without knowing which model made it.
- Your laptop becomes a screen for your desktop: pair it with a code and browse the library from another room, over an encrypted link on your home network.
- Make it yours: light or dark theme, orange or white accent, and in Aspetto (Appearance) you choose how each page looks and moves, with the little photo mascot keeping you company.
- Uses an NVIDIA graphics card when there is one, and works without it.
- Updates itself.
A companion phone app, which will show your library offline, is in development.
Install in three steps
- Download the installer (38 MB).
- Double-click it. Windows may say "Windows protected your PC" because the file is not signed yet: click More info → Run anyway.
- Open Local AI Frame Classifier from the Start menu. On the first start it prepares its AI engines for your hardware (a download of about 0.5 GB, or 2.7 GB with an NVIDIA card) and then guides you through choosing your photo folder.
No administrator password, no Python, no command prompt. Full installation guide →
Requirements
| System | Windows 10 (1809 or later) or Windows 11, 64-bit |
| Graphics card | optional; NVIDIA cards are used automatically |
| Disk space | about 2 GB, or 6 GB with an NVIDIA card, plus the models you choose |
| Internet | on the first start, and to download models; afterwards it works offline |
Your privacy
Photos, videos, labels, categories and names never leave your computer. The program connects to the internet only for these four things, and none of them carries your data:
| When | Where | What for |
|---|---|---|
| first start | pypi.org, download.pytorch.org |
download the AI engines |
| when you click the button | huggingface.co |
download the AI models |
| at start-up | api.github.com |
check for a new version |
| when updating | github.com |
download the new version |
Guides
Help and feedback
Found a bug or have an idea? Open an issue. Never attach your personal photos, and never include the names of people or categories from your library: a description of what happened and the version number are enough.
To report a security problem privately, see SECURITY.md.
Licence
Copyright © 2026 Lorenzo Dainelli. All rights reserved. You may download and use the program on your own computers; see LICENSE.md. The source code is not public.
来源:GitHub · github.com