A Working Model project
Audiary is a simple, private text-to-speech app for iOS. Paste or share any text — articles, PDFs, EPUB books, notes, LLM output — and listen to it like a podcast. No accounts. No subscriptions. Everything stays on your device.
Audiary turns text into audio on your own device. The neural voices come from an AI model that downloads once and then runs entirely on your phone — no accounts, no per-word API bills, no waiting on a server.
Twenty neural voices that sound closer to a person reading than to a system alert. The model is a one-time ~90 MB download; after that it runs on your device's Neural Engine with no network connection at all. Apple's enhanced and default voices are there too if you prefer them, or want to skip the download.
Paste text directly or use the iOS Share Sheet to send content from any app. Notes, markdown files, selected text — it all works, and you can edit anything before it saves.
Share a web page and Audiary saves the article itself — not the URL, not the navigation, not the cookie banner. Section headings become chapters, footnote markers and reference lists are stripped, and links read as their text. Share from Safari and extraction happens right in the page; share a bare link from Mail or Messages and Audiary fetches just that one page.
Import any PDF from Files, or share one from another app. Books with a real text layer arrive in seconds, with chapters from the PDF's own outline. Scanned page images are recognized on your device, page by page, with progress you can pause and resume — no page is ever uploaded anywhere.
Import an EPUB from Files, share one from any app, or tap a book anywhere in iOS and choose Open in Audiary. Chapters come straight from the book, footnote markers and tables of contents are left out, and chapter numerals are spoken as numbers. Copy-protected (DRM) books can't be read — Audiary says so rather than failing quietly.
Lock your phone, switch apps — Audiary keeps playing. Full lock screen controls with play, pause, skip, and a working scrubber.
Group your audiaries into folders once the library outgrows a single list. Reading queue, research, work — however you sort things.
Markdown headers — and the section headings of imported articles — become chapters automatically. Skip between sections like tracks in a playlist.
Follow along in full-screen mode with real-time word highlighting as Audiary speaks. Good for focus and retention.
Teach Audiary how to say the names and jargon it gets wrong, and it remembers. Heteronyms are resolved from context, so a word like "record" comes out differently depending on whether you're setting one or making one.
Listen at your pace, from 0.75x to 1.5x, with your preference saved between sessions.
Every control — player, voice picker, library — is labeled and operable with VoiceOver. Dynamic Type is supported throughout.
Audiary has no servers, no analytics, and no accounts. Every voice — neural, enhanced, and default — synthesizes speech on your device. We never see your text, because there is nowhere for it to go.
No data collection or tracking
No accounts or sign-ins required
All speech synthesis on-device
Network use is limited to the one-time neural model download and fetching a web page you explicitly share — no text or identifiers, ever
PDFs and EPUB books are read on-device — scanned page images are never uploaded
Crash reports stay on your device unless you choose to email them
No microphone or camera access
Uninstall deletes everything
Try free for 15 minutes. Pay once if you want unlimited listening. That's the whole pricing model.
Bug reports, feature requests, real questions — they all land in my inbox. No queue, no ticket system, no bot.