ainewsblitz.com

Breaking

AI File Sorter Adds Review-First Local AI Renaming and Folder Organization

  • Open Source

AI File Sorter is a cross-platform desktop app that uses local or remote language models to suggest clearer filenames and organize cluttered folders, while leaving the final changes under user control. The project supports Windows, macOS, and Linux, and its latest 1.9.2 release includes a Windows startup fix, following a broader 1.9.0 update that added file previews, smarter category whitelists, safer recursive scans, custom visual models, and configurable model storage.\n\nThe core workflow is content-aware rather than limited to filename or extension rules. The app can inspect supported images with a visual language model, extract text from documents, and use embedded audio or video metadata to propose normalized filenames. A generic image such as IMG_2048.jpg can receive a descriptive suggestion such as clouds_over_lake.jpg, while supported documents can be renamed based on their contents. Categories and optional subcategories are also proposed for folders such as Downloads, external drives, archives, and NAS storage.\n\nThe process is deliberately review-first. AI File Sorter presents proposed destinations, categories, and filenames in a review table before applying anything. Users can edit or reject suggestions, run a dry preview, choose a rename-only workflow, and use persistent undo to reverse the latest sort. The project describes the application as free and open-source, with local inference available without an internet connection or API key.\n\nLocal processing is built around models including Gemma 3 4B IT, Gemma 1.1 7B, and Mistral 7B, as well as user-provided GGUF models. Gemma 3 4B IT is the default visual backend, while LLaVA 1.6 Mistral 7B is another supported visual option. Image analysis requires both the main model file and a matching mmproj projector file. When local models are used, the project says files, filenames, images, and metadata remain on the computer and no telemetry is sent. Users can instead connect their own OpenAI or Gemini API keys, or a custom OpenAI-compatible endpoint, when they want remote inference. Details on the supported local and remote workflows are documented in the project repository and the official FAQ.\n\nDocument analysis covers PDF, DOCX, XLSX, PPTX, ODT, ODS, ODP, and common text formats. Audio and video rename suggestions can use embedded tags such as ID3, Vorbis comments, and MP4-style metadata. The application also offers category whitelists, including category-specific subcategories, so users can constrain the vocabulary used for a particular cleanup session. Its consistency cache stores prior results and approved review decisions locally; the documentation says this state guides later suggestions but does not train or modify the underlying model.\n\nThe recent release history shows a gradual move toward safer automation. Version 1.5.0 introduced local image analysis and rename suggestions, 1.6.0 added document content analysis, 1.7.0 added metadata-based audio and video filenames, and 1.8.0 made Gemma 3 4B IT the default visual model while adding local learning from approved reviews. Version 1.9.2, released on August 14, 2026, fixed an issue in which the Windows app could close before sorting began, according to the official changelog. An optional Windows Explorer Extension released on August 20 starts categorize and rename workflows from the File Explorer context menu; it requires AI File Sorter 1.9.2 or later and is sold as a paid convenience add-on, while the main application remains free and open-source.\n\nHands-on use has highlighted both the appeal and the trade-offs. A recent evaluation found the local visual analysis useful for identifying screenshots and other image content, and considered the review table, dry-run mode, and undo feature strong safeguards. The same evaluation found that categorization was generally more reliable than rename precision: some outdoor images still received generic names. Processing can also be slow when visual analysis is applied to a large folder, particularly on the CPU; the FAQ notes that a discrete GPU with around 6 GB or more of video memory can make image scanning much faster.\n\nThere are practical coverage limits as well. Legacy binary Office formats such as .doc, .xls, and .ppt are not supported, while modern Office XML and OpenDocument formats are. Visual analysis needs two model files, and systems with limited video memory may fall back to slower CPU processing. The FAQ recommends working with focused folders or smaller batches rather than sorting an entire disk at once, and the application does not currently include a dedicated low-resource background-throttling mode.\n\nAI File Sorter’s significance is less about autonomous file manipulation than about applying multimodal AI to a task that has traditionally depended on manual inspection or brittle scripts. Its current design keeps the model responsible for interpretation and suggestions, while the user remains responsible for approving file operations. That division makes the tool useful for messy personal archives, downloads, documents, images, and media libraries without requiring users to surrender control of the filesystem.