Cleans the transcript
Fixes recognition errors and punctuation, removes fillers and accidental repeats, and leaves the facts alone.
transcriptAI dictation for iPhone
AiDdictEd turns speech into finished text in a normal text field on iPhone. Pick a style or another language while you record; the keyboard applies it when you finish.
The public beta is on TestFlight. Joining is free, and the build is early, so feedback matters.
Icon 2.0 · project mark
AI cleans the transcript. Dict recognizes speech. Ed changes the tone or language.
Fixes recognition errors and punctuation, removes fillers and accidental repeats, and leaves the facts alone.
transcriptRussian, English and Italian in normal text fields: notes, messages, email, documents.
speechFour styles and translation. The facts stay; the tone and language change.
wordingStart recording, then tap a style or a language. The choice is fixed when you finish, so the text already matches what you need.
Well, anyway, I looked at the file, and, um, it’s not that bad, but I think we missed the deadline after all.
Grey text is removed by editing.
I looked at the file. It’s not that bad, but I think we missed the deadline after all.
Meaning kept, fillers removed.
This is an example. In the app you choose the style and language while recording, and the choice is applied when you finish.
Styles other than word-for-word use on-device text editing. If it isn’t available for the language, the full transcript stays as it is.
It takes a couple of minutes. The keyboard sits next to the system keyboard; switch with the globe key.
Install TestFlight, open the link and install AiDdictEd.
Settings → General → Keyboard → Keyboards → Add New Keyboard → AiDdictEd.
It’s needed for voice input, inserting text and shared history. Typing works without it.
Open a normal text field, tap Start and speak. Change the style or language while recording.
Screenshot from a test build. The interface is still in English.
Swipe, emoji and history are in the same keyboard. The language switch is there too.
75,000 words per language. Slide over the letters and the word is inserted with a space.
Categories and 32 recent. Compound emoji are inserted whole.
Recordings from the app, right in the keyboard. Tap a line to insert its text.
Download the model and dictate without a network. Or switch to cloud mode when you need it.
Personal settings and the word count move to a new device.
Key sounds, haptics, layout and word count in the keyboard panel.
RU · EN · IT
Switch layouts as you type. Translation needs a language pack, which you download once.
Четверг, десять утра, встреча в офисе. Не забудь документы.
RussianThursday, ten in the morning, meeting at the office. Don’t forget the documents.
EnglishGiovedì, alle dieci, riunione in ufficio. Non dimenticare i documenti.
ItalianTranslation works after the language pack is downloaded. Editing needs on-device text processing that supports the language.
They don’t show up in screenshots. They make recording less annoying.
The timer stops and the pause isn’t in the audio or the transcript. Time to think.
If translation or editing fails, the full transcript stays. An empty answer never replaces it.
Words are counted before translation. Re-processing an entry replaces it instead of adding it again.
It doesn’t start when the app opens or when you come back. Only when you tap it.
Privacy
The keyboard handles what you type, so the mode matters. Choose what stays on the iPhone.
Recognition runs on the phone after the model is downloaded. No internet needed.
Audio is sent for transcription. If cloud cleanup is on, text can go too.
Translation runs on the iPhone after the language download. Editing runs when on-device processing supports the language.
Full Access is an iOS permission for voice and inserting text. It doesn’t turn on cloud processing by itself.
If something is missing, send feedback from TestFlight. You can attach a screenshot there.
Early beta. Some things will break. That’s how we find out what to fix.
It’s a build before the public release. The main parts work, but features can change. We’re collecting testers and feedback.
Install TestFlight, open the beta link and install AiDdictEd. Then add the keyboard in Settings → General → Keyboard → Keyboards → Add New Keyboard → AiDdictEd. You need iOS 17 or later.
iOS needs it for voice input, inserting text and sharing history with the app. Typing works without it. It’s a system permission, not a cloud subscription.
In offline mode, recognition runs on the iPhone after the model is downloaded. In cloud mode, audio is sent for transcription. Cloud cleanup can send text if you turn it on. Translation runs on the device after the language download; editing runs when on-device processing supports the language.
Russian, English and Italian. Dictation handles mixed speech. Translation is available for all three after the language pack is downloaded. Swipe uses the layout you selected.
Yes, if the offline recognition model is downloaded. Turn off cloud cleanup if you want to keep the whole workflow on the device.
The current build is in English. Dictation, editing and translation support Russian, English and Italian; the styles are called Word for word, Clean, Polished and Formal.
No. AiDdictEd is an additional keyboard. Switch with the globe key. Swipe, emoji and history are built in.
Early beta · Testing
Install TestFlight, open the link and add the keyboard. It’s free. The build is early, so tell us what breaks.
Early beta. Features and behavior can change.
AiDdictEd 2.0
TestFlight