Speech recognition
Turns the person's speech or audio files into text, possibly on Apple servers.
In one line
Can turn your speech into text, sometimes on Apple's servers
Turns the person's speech or audio files into text, possibly on Apple servers.
Reveals what is said near the phone. iOS warns that audio may go to Apple servers; on-device modes avoid that.
What it can and can't do
What the user does: You tap the microphone button and tap Allow for speech recognition
It can
- Turn your speech or audio files into text
- Use on device recognition on Android 13 and later, which stays on the phone
- Use Apple's server mode, with a limit of about 1 minute per request and a daily quota (iPhone)
It cannot
- Use it after you tap Don't Allow
- Listen in the background on Android 17 under tighter rules
- Keep the audio on the phone when server mode is used on iPhone
How it differs between platforms: Both ask first. iPhone warns that audio may go to Apple's servers. Android 13 and later offers an on device mode.
On Android and iPhone
| Android | iPhone | |
|---|---|---|
| Status | Supported | Supported |
| First version | Android 1.0 (API 1) | iOS 10 |
| Permission or entitlement | RECORD_AUDIO (no separate speech permission) | NSSpeechRecognitionUsageDescription (plus NSMicrophoneUsageDescription for live audio) |
| Protection level | dangerous | prompt |
| How access is granted | prompt | prompt |
| What the user sees | Allow microphone while using, only this time, or don't allow | Allow or Don't Allow speech recognition |
| How long it lasts | Until revoked | Until revoked |
| How to take it back | Settings > Apps > Permissions > Microphone | Settings > Privacy & Security > Speech Recognition |
| Store policy | Play: microphone foreground service rules | App Review 5.1.1 |
| API | SpeechRecognizer, createOnDeviceSpeechRecognizer (API 33) | SFSpeechRecognizer, SpeechAnalyzer (iOS 26) |
| Notes | Android 13 added an on-device recognizer; Android 17 tightens background audio | Server mode limited to about 1 minute per request and a daily quota; SpeechAnalyzer is on device |
Data you get
What comes back: Transcription with confidence
Shape: stream
How often: streaming
| Field | Type | Unit | Range | Precision |
|---|---|---|---|---|
maxDuration | integer | seconds | about 60 per server request on iOS | Not listed |
Technical limits
- Separate mic permission is also required for live audio
- Server limits about 1 minute per request
- Language pack limits
The story
What it is
Turns your speech or an audio file into text, sometimes by sending it to Apple's servers.
For you
A pop up says speech might be sent to Apple. After you allow it, you can dictate notes or search with your voice.
For makers
It offers dictation, captions and voice commands without building models. Server mode is stronger but sends audio off device, while on device mode keeps it private.
Where you see it
| Type | Example | Apps |
|---|---|---|
| productivity | You dictate a note and see text appear as you talk. | Notes, Otter |
| video | A video app adds live captions for a hard of hearing viewer. | Zoom, YouTube |
| games | A word game lets you speak your guess. | Heads Up |
| education | A language app listens to your accent and scores it. | Duolingo, Babbel |
| accessibility | Voice control types and taps for someone who cannot use their hands. | Voice Control |
| smart-home | You say a command and the app turns it into an action. | Apple Home |
| messaging | You dictate a text message while driving. | Messages |
Risks
- Reveals what is said near the phone
- Audio may go to servers
Android and iPhone in short
- Android: Android's SpeechRecognizer may use Google servers, and offline packs can be downloaded.
- iPhone: iOS warns that audio may go to Apple, and on device mode avoids that.
What the platform says
- Apple: Speech framework. Convert audio to text with on device or server based recognition, after the person agrees.
Good uses and abuse
Legitimate uses
- Dictation
- Voice commands
Known abuse
- Covert transcription
Scores
| Score | Value | Meaning |
|---|---|---|
| Sensitivity | 4 of 5 | How revealing the data is |
| Utility | 4 of 5 | How useful it is to honest apps |
| Friction | 2 of 5 | How much the user has to do |
| Fingerprint bits | Not rated | Rough identifying entropy, when known |
History
- 2016-09, iPhone iOS 10: Speech framework and prompt
- 2022-08, Android 13 (API 33): On device speech recognizer added
- 2022-08, Android 13: On device speech recognizer API
- 2025-09, iPhone iOS 26: SpeechAnalyzer for on device long form transcription
- 2025-09, iPhone iOS 26: SpeechAnalyzer
Related capabilities
Related
- Recording indicator dot. Can't use the camera or mic without a green or orange dot showing
- On-device translation. Can translate text on your phone, after you download a language pack
Same idea on the web
- Web Speech recognition. Can turn your spoken words into text while you speak and the tab is open
Sources
- Apple: Asking permission to use speech recognition
- Android: SpeechRecognizer
- Apple: Protected resources keys
- Apple: SpeechAnalyzer
- Android: createOnDeviceSpeechRecognizer
How well checked: Based on official documentation.