Dictation
Speak into the field you are already in. Most of it runs on the phone, and the part that sometimes does not is worth knowing about.
Clink for iOSStart dictating
Speak to type. Your words are transcribed on-device and streamed straight into whatever field you're typing in. Adds a mic to the panel button. In the next iOS and Android test builds, Dictation settings open on an overview with a pinned preview. Enabled, Show in toolbar and Show in Tools menu sit together at the top. Each section has a short description of its controls. Choose Look for recording styles, the glass capsule and colors, or Placement for where dictation appears. Tap Dictation in the heading to return. The iOS overview also has AI for Apple Intelligence cleanup; Android does not offer Apple Intelligence.
- Open the tools button on the suggestion bar.
- Tap the microphone.
- Talk. The words arrive in the field as you go, and you can type around them at any point.
In the next test builds, inline recording uses the waveform and stop control without an extra red dot beside the waveform. Tap the waveform to finish the current dictation. In the next iOS test build, tapping the microphone on Home first explains dictation and that the microphone can stay on after you leave Clink. Choose Start dictating to activate it, or Cancel to leave it off. This explanation appears each time you start from Home. While a session is active, tap Dictating on Home to turn the microphone off immediately. In the next Android build, closing dictation while the speech service is still starting cancels the pending microphone request. If dictation has already begun, Clink gives the service time to return your final words.
What it needs
Settings path: Home Dictation
The microphone and speech recognition, both of which iOS asks for the first time you use it. Dictation is one of the Membership tools.
How much of it is offline
Dictation uses on-device SpeechTranscriber when available on iOS 26 with its language resources ready. The compatibility speech recognizer can send audio to Apple’s servers, including while those resources download. Dictation is not always offline. In the Android test build, allow microphone access when you first open dictation. Clink tries on-device recognition first. If the selected language is unsupported or its speech pack is missing, Clink tries the phone’s default speech service. That service may send audio to its provider and need an internet connection. After a temporary connection or speech-service error, Clink keeps the words already typed and makes up to two recovery attempts. Wait for listening to resume before continuing; speech during an interruption may not be captured. If dictation still fails, check your connection and the speech service’s language downloads, then tap to try again.
Dictation assist
Speech recognition gets the words right and the punctuation wrong. When you stop speaking, the model puts the sentence breaks and capitals back, and drops the "um"s.
Without opening the app
A home screen widget starts a session, and the session itself runs as a Live Activity you can end from the Lock Screen.
Swap the whole app for a single dictation screen, and the keyboard opens on the mic instead of the keys. Nothing is switched off: the keys are one tap away, and every setting you have keeps working. On iPhone, returning from dictation keeps your saved custom theme, including edits to individual keys. In test builds, the style choices in Dictation settings are still previews. Select one to see it animated in the main preview.
Optional Wizard cloud voice
Wizard cloud voice is coming soon. It is off by default and separate from free, unlimited on-device dictation. When it becomes available, open Dictation settings or the Clink Wizard page and enable Use Wizard cloud voice. A current Wizard subscription and a private Clink account are required.
Cloud voice sends dictation segments of up to 60 seconds through ANTI and Vercel for transcription. Longer segments, offline use and cloud failures keep the local result. Audio is not saved for a later upload. Offline recognition needs a local speech model for the language you are using. On Android, cloud voice requires Android 13 or later and an available on-device recognizer.