openaver

OpenAver is a free, open-source desktop application built around a "browse by cover, find by actress" approach to collections, with optional output of the NFO and cover images required by Jellyfin/Emby/Kodi. The project is released under the MIT license and emphasizes 100% local execution, no account registration, and no uploading of file information; network requests are used only to scrape publicly available metadata. Installation and platforms It supports Windows 10/11 and macOS (Apple silicon M1 and later). A one-line install script is provided (curl for macOS, PowerShell for Windows), and you can also download a ZIP from Releases or a .bat installer for Windows. After installation, everything is done through a graphical interface, with no Docker and no commands to type. There is no official installer for Linux, but the documentation explains that you can host it yourself as a LAN server and use it through a browser. Three main pages 1. Scanner: Add the folders containing your videos and click "Generate Webpage" to import them. If NFOs and cover images already exist alongside the videos, the existing data is read directly, and the scanning process is read-only. After scanning, items missing an NFO or cover are listed, and these can be filled in from the internet with one click; when filling in, only empty fields are populated and existing data is not overwritten. 2. Showcase: Displays the collection as a cover wall, with stackable filters. Actresses have their own separate wall, with data cards providing height, cup size, measurements, age, alias history, and more, and sorting by cup size, age, height, or number of titles. There is also rule-based similar-title recommendation (tags, series, studio, actress), performed locally and instantly offline. 3. Search: Handles newly downloaded videos. You can drag in files or folders, and it automatically recognizes the product code, queries data in batches, and downloads covers and stills. After confirming, click "Organize" to rename, create folders, and write NFOs. There is also a bookmark feature and optional scheduled organizing (automatically processing specified folders every 12 hours). Scraping sources There are 8 built-in sources: JavBus, Jav321, JavDB, DMM, D2Pass, HEYZO, FC2, and AVSOX. They can be queried simultaneously, compared against your library, and marked as already collected. There are also two backup sites, JavLibrary and FC2-javten, which sit behind Cloudflare human verification and therefore only support manual exact product-code queries in the desktop version. Advanced users can host Metatube themselves and connect it to the federation, bringing the total to more than 30 community-maintained providers. Integration with playback software When organizing, you can choose to output the poster and fanart filenames and NFOs recognized by Jellyfin/Emby/Kodi, while leaving the videos in place. For collections you do not want modified, such as those on a NAS or cloud mount, you can set them as "read-only" sources: the original files are not written to at all, and the fetched data and covers are output to a specified local folder. You can also generate .strm files so playback software can stream the original files directly; cross-machine path replacement rules are also supported. Other features - Actress and tag aliases: You can add aliases in the interface, and searches automatically expand to Chinese, Japanese, and English synonyms, merging the stage names and retired names of the same person into a single card. - AI translation: Supports Ollama (free locally), Gemini Flash, and OpenAI-compatible endpoints, and can translate Japanese titles into the interface language. - Mobile browsing: In settings you can switch to "server" mode, letting phones or tablets on the same Wi-Fi browse via a browser. It is not exposed externally by default, and a password can be set. - Static HTML export: Can generate a standalone HTML file for offline browsing. - Four interface languages with instant switching: Traditional Chinese, Simplified Chinese, Japanese, and English. AI operation API A local REST API and capabilities manifest are provided. AI tools (such as Claude Code, Cursor, etc.) can use a single curl command to read the endpoint descriptions and then connect multi-step operations themselves, for example batch-favoriting actresses, creating aliases, batch-adding tags, or generating an HTML page with covers. Endpoint definitions are centralized in web/routers/capabilities.py. Technical architecture The backend is FastAPI (Python 3.12), the frontend uses Jinja2, DaisyUI, Tailwind CSS, and Alpine.js, the desktop wrapper is PyWebView, the database is SQLite (WAL mode), and tests are written with Pytest. The project documentation explains how to run development mode and desktop mode from source, and provides packaging scripts for Windows and macOS. Privacy and file safety The project states that it does not upload videos or library lists, and that there are no accounts or telemetry. Searching, browsing, and scanning are all purely read-only; filling in data only adds NFOs and covers alongside the videos; only when the user actively clicks "Organize" are videos moved or renamed according to the rules, and files are not deleted. If a file with the same name already exists at the target location, you are warned first. License and community Released under the MIT license, with GitHub Issues and a Telegram group provided as channels for reports and discussion. The project also includes a disclaimer reminding users to respect website terms of service, keep request rates reasonable, and use it only for personal learning and research.