PDF AI Renamer 2.0 is out — full redesign + on-device analysis for docs & images
Following up on my earlier posts here — 2.0 just shipped with a full redesign and on-device OCR for scans/images. Posting per the new PCP format: Problem PDFs and scans pile up with meaningless names — scan_0042.pdf, IMG_3891.HEIC, Unbenannt-3.pdf — and manually opening, reading, and renaming each one doesn't scale. Existing "smart" tools either rely on rules based on metadata you already have to know (useless when the filename itself carries zero information), or they ship your document content to a cloud API — a non-starter for invoices, contracts, or anything sensitive. PDF AI Renamer reads what's actually inside the file — text or, as of 2.0, scanned images via on-device OCR — and generates a structured filename from it, entirely on your Mac via Apple's MLX framework. No server round-trip, no text layer required. Comparison - Hazel — the classic macOS automation tool, but purely rule-based: it moves/renames based on metadata, dates, or folder location, not document content. It can't tell that scan_0042.pdf is an April invoice from a specific vendor — it has no way to read that out of the file. PDF AI Renamer is content-aware: it extracts sender/date/subject from the actual document (OCR'd if needed) and builds the name from that — no rule maintenance, no folder-watcher setup, works the moment you drop a file in. - RenameClick / Zush — closer competitors: both also do AI content-based renaming for PDFs and images. But their free/local tiers are capped (e.g. a monthly credit allowance) before you're pushed toward cloud AI or a subscription just to keep renaming. PDF AI Renamer's inference runs on-device with no per-rename cloud dependency — since 2.0 this now also covers scans and images (JPEG/PNG/HEIC/TIFF) via on-device Vision-framework OCR + a vision-capable local model (Gemma3 4B), not just PDFs with a text layer. Example: drop in scan_0042.pdf (an unreadable scanned invoice) → OCR'd and read on-device → renamed to something like 2024-03-15_Acme-GmbH_Invoice_1234.pdf, using your own custom template and placeholders. Pricing Monthly subscription: 2,99€ Annual subscription: 17,99€ Lifetime: 19,99€ Product & developer websites: Apple AppStore (MacOS) https://pdf-ai-renamer.com https://alexander.giehoff.de
Collected discussion
Would it be possible to automatically detect new PDFs, run them through this, and propose the change? Could be cool to make it proactive!
Yes, I think that is exactly what the import folder does. You can define a folder to be monitored for new files. When a new file is added, it is automatically loaded into the queue. Currently, you have to trigger the suggestions with a single click. Extending the app to include automated renaming is certainly a possibility if there is a demand for it.
Nice! I especially like the on-device approach for privacy. It would be interesting to see batch rules or custom naming templates evolve further for different workflows.
Thanks for your feedback! Can you explain how you would like to use them in different workflows?
The missing trust layer for me is not more automation; it is evidence attached to each proposal. For 2024-03-15_Acme-GmbH_Invoice_1234.pdf, show which OCR spans produced the date, vendor, document type, and number, and flag fields the local model could not ground. A watched folder should default to queueing proposals, keep the original name and path in history, and make bulk approval reversible. That gives users a path from review-first to automation after a template proves itself, without requiring blind trust on day one.
Hi and thanks for your feedback! I am already thinking about a new feature, that makes the results of the model more transparent. Therefore I am happy to hear, what your expectations are. Regarding the import folder there is actually no „blind“ transformation. New files are added automatically to the queue. Then you start the renaming process manually and you are able to review the results. To enhance this process with even more automation - on a „trusted way“ - is a feature at the backlog, too. If you like to share more details of your expectations for these features, please let me know.
goodluck w sales
PDF AI Renamer — by Alexander Giehoff Ratings: 0 (no ratings yet) · 4 reviews Average: n/a ★ Age: released ~12 months ago · rated 4+ Overall score: 0 / 100
nice
Hazel excels at reading indexable PDFs. What's more, the latest updates have introduced an automatic OCR feature that can read non-indexed PDFs. Previously, all you had to do was use a ruler to OCR a PDF before processing it. I've been doing this for many years and get better results than with an AI assistant that generates random names, while Hazel performs my tasks exactly as I want them.
just curious, what will motivate people to use this app if they have chatgpt or claude subscription
Privacy.