Questions that come in.

Yes. Yappy is open source under the MIT licence. Being able to hear what you read is an accessibility matter, and accessibility tools should not ask for a credit card. On the desktop there is no paid tier and no ads. On the iPhone, the free parrot carries three documents at a time on its perch, with no time limit; if you want a bottomless perch, the audiobook press and the bridge to your computer, you wind it up with Yappy Parlanchín, as a subscription or a single lifetime purchase. No account and no ads either way.

There is no catch: there are no servers to pay for. Synthesis runs on your device, so Yappy has no per-user cost. It is a personal project by its author, published because it already existed and someone else might need it.

No, on purpose. The three-document perch is the trial, and it neither expires nor turns into a charge after seven days. If Yappy is useful to you, you will know without anybody holding a stopwatch.

Yappy's code is MIT and puts no conditions on you. The voice model, Supertonic 3, has its own: an open OpenRAIL-M licence with use restrictions attached. Check it on the model's page before any commercial use, because that part does not depend on Yappy and we cannot promise on its behalf.

No. Synthesis, transcription and article extraction all happen inside your device. What you share with Yappy is not sent anywhere: there are no accounts, no copies, and no Yappy server where they could be kept. Four specific things do touch the network, and all four are listed in the privacy policy.

Yes, it does, and that is worth saying without hedging. The mobile app sends anonymous usage statistics: they come switched on and you can switch them off in the back room. What travels is the name of an event from a closed list (the app was opened, which onboarding step is on screen, which plan was tapped, whether a reading finished), a random installation identifier, the app's language and its version. Never your name, your email, the text you read, a document title or a link, and no advertising identifier. And it does not go to Google, or Firebase, or any analytics company: it goes to a Cloudflare Worker and a D1 database belonging to the author himself. On the desktop there are no statistics of any kind.

It is a direct connection between your two devices, encrypted end to end over QUIC, using iroh's protocol. Pairing happens by pointing the camera at a QR code your computer shows, and the resulting key can be revoked from there at any time. There is no Yappy server in the middle, because no Yappy server exists.

Yes, forever. After the initial model download — about 200 MB on the phone, about 398 MB on the computer — you can use Yappy in airplane mode: paste text, open documents, synthesise, export. Only sharing new links needs a network, for obvious reasons. There is no cloud sync either: your library lives on your device, and the bridge is the closest thing to moving something from one machine to another.

Supertonic 3, an open model by Supertone of about 99 million parameters that fits in 398 MB on the desktop and 200 MB on the phone, and synthesises faster than real time on any laptop from the last few years. It works at character level: that is why the reading script exists, turning numbers, acronyms and symbols into words before the voice ever sees them.

Arabic, Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Latvian, Lithuanian, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Turkish, Ukrainian and Vietnamese. Thirty-one of them. And all ten voices are polyglot: the same voice reads every one without changing timbre.

Yappy detects the document's language and, paragraph by paragraph, each piece's: an English quotation inside a Spanish essay is read in English, in the same voice, without you touching anything. And if detection gets it wrong, you can set the language by hand.

System readers read the screen as it is, with the system's prosody and no memory. Yappy prepares the text before saying it (cleans the article, verbalises numbers and acronyms, marks the pauses), reads it in a natural voice, remembers where you were, exports audiobooks and keeps playing with the screen locked. Cousin tools, different trades, and they work together without getting in each other's way: Yappy does not replace a screen reader and does not try to.

Web articles (the page cleans itself up), videos via their published transcript, PDF, EPUB, Word, ODT, RTF, Markdown, plain text, .pptx, .xlsx, .csv and voice notes. On mobile, everything through the share sheet; on desktop, also by dragging or with the shortcuts. And if you have pandoc installed on the computer, the list grows to LaTeX, reStructuredText, Org and a few more.

On the desktop they go through a local OCR before being read, and nothing leaves your computer there either. On the iPhone, not yet: if the PDF is a photograph of paper, the phone finds no text to read. A crooked or blurry scan gives what any OCR can give, but a decent one reads through whole.

Only the ones you can open. An unprotected EPUB reads through whole, chapters included. Books with a store's DRM cannot be read, here or anywhere else: that lock is not ours, and we are not going to pretend we have the key.

From the library itself: every document has its forget option, and the audio goes with it. Exported files (.m4b, .mp3, .wav) are yours and live wherever you saved them; deleting or moving them is a job for Finder or Files.

The model loads into memory the first time. After that it stays warm and the following sentences come out on the fly. If you have just installed the app on a computer, that first wait also includes downloading the voice model.

There is no user pronunciation dictionary, so the quick route is to spell the word the way it sounds in the editor: the model reads characters and will do as it is told. If it is a general pattern (an abbreviation, an acronym, a unit), say so in the GitHub issues: the reading script's tables grow from cases like that.

Yappy takes the video's published transcript; it does not transcribe its audio. If that video has no transcript, there is nothing to read. And if the video is yours and you have the file, share the file instead: then the local ear transcribes it.

In the GitHub issues, or by writing to [email protected]. With the text that failed, if you can share it, and your system, it gets fixed sooner.