WordCatch
Teach it once. It keeps learning.
They repeat a word until someone says it back. WordCatch listens for that moment and learns from it, so the family teaches it just by talking the way they already do.


Built for one family. Ready for anyone.
Conversation mode
Talk as usual while their words appear as sentences and are said out loud. Words said one after another become one sentence.
What next?
After their word, big picture buttons offer what they usually say next, like “want” → cookie, outside, iPad. One tap finishes the sentence, and every tap teaches it their habits.
Finds new words
Sounds it doesn't know yet are kept. When the same sound keeps getting the same answer, it suggests the new word, so their vocabulary keeps growing.
Knows their voice
A voice model on the device tells them apart from family and caretakers, so it only ever learns from them.
Speaks for them, naturally
“cookie” becomes “I want a cookie, please,” in a natural voice. Sentences are made on the device, free.
Works with their AAC
Bring in their vocabulary from Proloquo2Go screenshots, a word list, a board file or a PDF, with pictures from free symbol sets.
Progress for their team
A report for their speech therapist or teacher, as a PDF or spreadsheet, with how they say each word and how well it's understood.









One app. A lot going on.
Everything WordCatch does runs on the iPhone or iPad. The core app stays small, and the bigger models are optional downloads you choose, see, and can remove any time.
Learning
- Learns from real momentsWhen they repeat a sound until someone says the word back, WordCatch learns what it means.
- Knows when it doesn't knowA sound unlike any word it has learned shows as “?” instead of a wrong guess, and is kept so it can learn it.
- Finds new wordsThe same sound answered the same way a few times becomes a suggested new word.
- Gets better overnightRetrains while charging, tries several setups, and keeps a change only if it's more accurate on recordings a person confirmed.
- Sound spellingsOptional: listens to each word's recordings and suggests how it's said, like “gah-ee” for cookie.
Speaking
- Full sentences“cookie” becomes “I want a cookie, please,” written on the device with Apple Intelligence, or an optional model on devices without it.
- Natural voicesHuman-sounding voices you can hear before downloading, or the voices already on the device.
- Conversation modeTalk as usual; their words appear as sentences and are said out loud, like a translator app.
- What next?Big picture buttons for the words they usually say next, so one tap finishes a sentence.
- Ball or bath?When two words are close, it asks instead of guessing, and learns from the answer.
Pictures and words
- Find a SymbolEmoji and over 2,000 of Apple's symbols, offline, plus free AAC symbol sets: Mulberry, PiCom, OpenMoji and Stellar.
- Your own photosTake a photo of the real cup or cookie; the background can be removed on the device.
- Bring in their AAC wordsFrom screenshots of their AAC app, a PDF of a board, a word list, or a board file.
- Clean recordingsTrim a recording to just the word, or split one with several words into one each.
Family and team
- Knows every voiceFamily and caretakers read a short passage once; only a voice fingerprint is kept, never the recording.
- More than one deviceRecord on a second phone or iPad and send it with AirDrop; the main device adds it by itself, without duplicates.
- Progress for their teamA report for their speech therapist or teacher, as a PDF or spreadsheet.
- Help on every screenPlain-language instructions behind the “?” on each screen.
Storage and battery
- Half the space, same soundRecordings are stored in Apple Lossless: about half the size, with every sample checked to be identical.
- Archive, never deletePast 30 recordings of a word, extras can move to your backup folder, only when recognition is just as accurate without them. Restore any time.
- Backups you controlAutomatic backups to a folder you choose, such as iCloud Drive.
- Easy on the batteryBackground listening pauses in Low Power Mode, on low battery, or after a quiet stretch.
Private by design
- On the deviceRecordings, recognition and sentences stay on the iPhone or iPad. No accounts, ads or tracking.
- Only their voice is keptOther people's speech becomes text on the device to spot the moment they're understood, then it's gone.
- Diagnostics you shareAn optional text-only log of guesses and answers, to help improve WordCatch. It leaves the device only if you send it.
Downloads you choose
Each one is optional, runs entirely on the device, and can be removed in Settings to free the space.
| Download | What it adds | Size |
|---|---|---|
| Speech and voice models | Hears when they finish talking and tells voices apart. Downloaded once, automatically. | about 20 MB |
| Whisper | More accurate recognition and speech-to-text, measured on their own recordings before it's used. | 75 or 150 MB |
| Natural voices | Human-sounding voices, previewed before you download. | 124 MB, then about 0.5 MB each |
| Sentence model | Full sentences on devices without Apple Intelligence. | about 700 MB |
| Sound spellings | Suggests how each word is said, from their recordings. | about 300 MB |
It stays on the device.
Only their voice is kept
Other people's speech is turned into text on the device to spot the moment they're understood, then thrown away.
The AI runs on the device
Speech detection, voice recognition, their personal model and sentences all run on the iPad or iPhone. No audio is ever sent anywhere.
No accounts, no tracking
No ads, no analytics. Backups go to a folder you choose, such as iCloud Drive. An optional diagnostics log is text only and stays on the device unless you share it.
Coming to the App Store.
WordCatch is being tested at home right now, with the family it was built for. It runs on iPad and iPhone, works alongside AAC apps like Proloquo2Go, and everything it hears and learns stays on the device.
hello@wordcatch.app
Made by ColeCustomWorks.

