Sound
A word you have only ever read is half learned
Rayaad speaks both sides of a card in the language that side is actually in. The voice comes from the language tag on your pair, not from your phone's default — so the language you are learning is not read out in the accent of the one you already speak.
- Both sides, each in its own language.
- Play on demand, or automatically as a card appears.
- Stored on the device, so it plays again with no connection.
The word you are learning
What it means
- Both sidesThe word and its meaning, each in the language it is written in.
- One voiceThe same voice for a language every time, so nothing shifts between cards.
- OfflineA clip you have played is kept on the device and works with no connection.
A sample clip, not the app's own voice. Rayaad generates real speech in the product; this page plays a static file so it never calls a metered provider.
Where sound shows up
Three places, one clip
The same cached audio serves all of them, which is why hearing a word a hundred times costs exactly as much as hearing it once.
In review
A speaker on the card, or automatic playback as it appears. Answering never waits for audio.
In listen mode
The due queue as a playlist: word, pause, meaning, next. Lock-screen and headphone controls, and it counts as review.
Rayaad on a phoneIn the extension
A word swapped into a page you are reading can speak itself from the popover, using the same cached clip.
The browser extensionThe mechanism
One clip per word, ever
A clip's name is a hash of three things: the exact text, the language tag, and the voice. That has three consequences worth knowing — replaying is free, editing a card gives you new audio rather than stale audio, and two learners studying the same catalog word are served the same file rather than paying for it twice.
- First playSynthesised, and counted against the day's allowance.
- Every play afterFree, instant, and available offline.
- Catalog audioMade once at publish time. Free on every plan, always.
- Edited cardDifferent text, different clip. Never the old pronunciation.
Why the language tag matters more than it sounds like it should
Most apps that speak a flashcard hand the text to the device's speech engine and let it guess. The guess is made from the characters, so a short word can be read in the wrong language entirely, and a word that exists in two languages is read in whichever the device prefers. The result is a pronunciation that is confidently wrong, which is worse for learning than silence.
Rayaad never guesses. Each side of a card belongs to a language pair, and the tag on that side is what the speech request carries. A language with a voice configured for it gets that voice; one without falls back to a default rather than to a guess.
What Rayaad does not claim
It does not claim native-speaker recordings — the audio is synthesised, and synthesised speech is very good at common words and less good at rare ones and at names. It does not claim every language is covered, because coverage follows the speech provider. And it does not offer pronunciation scoring or speech recognition: Rayaad does not listen to you, and there is no microphone permission anywhere in the product.
Next: where audio sits in a review, orwhat counts against the daily allowance.
About pronunciation
Which languages can Rayaad speak?
Speech is requested for the language tag on that side of the card, so coverage follows the speech provider rather than a list Rayaad maintains. A language with no voice configured falls back to a default voice rather than failing silently, and Rayaad never guesses a language from the text.
Does audio work offline?
A clip you have played before does, yes — it is cached on the device, and on the phone apps it is kept for offline listening. A word that has never been spoken needs a connection the first time, because that is when the clip is made.
Why is there a daily limit on audio?
Because synthesising a clip is a paid call. Only clips the server actually has to make are counted: replaying one that already exists is free on every plan, and so is catalog audio, which was paid for once when the unit was published. The Free plan allows 20 new clips a day and Pro allows 300.
What happens if I edit a word after hearing it?
You get new audio. A clip is named by a hash of the text, the language and the voice, so changing the text produces a different clip rather than new bytes at the old address. That is what stops an edited card playing its previous pronunciation.
Can I hear both sides of a card?
Yes, and each side is spoken in its own language — the one you are learning and the one you already speak are two different voices, chosen from the two tags on the pair rather than from your device default.
Twenty words tonight.
Free plan with no review limit, on every device. No ads, and your words stay yours.