Yes, both iPhone and Android can turn speech into text without an internet connection, but it depends on your phone, your language and what you’re transcribing. On recent iPhones, keyboard dictation runs on the device for many languages. On Android, Gboard offers offline voice typing for a number of languages, and Pixel phones do more of it locally. For recordings you already have, apps built on OpenAI’s open-source Whisper model transcribe entirely offline on most modern phones.
What you give up offline is mostly accuracy in hard conditions: noisy rooms, strong accents, rare languages and technical vocabulary.
Does iPhone dictation work offline? #
On many recent iPhones and in many languages, yes. Apple says your device will indicate in Keyboard settings whether your dictation audio and transcripts are processed on the device and not sent to Apple (Apple’s Siri and Dictation privacy page).
To check yours:
- Open Settings, then General, then Keyboard.
- Make sure Dictation is on, and read the note under the toggle and the dictation language list.
- If you use more than one keyboard language, check each. On-device support differs between languages on the same phone.
To test it: turn on airplane mode, switch Wi-Fi off too, open Notes and tap the microphone on the keyboard. If text appears, you’re dictating on-device. If you get an error, that language is using Apple’s servers.
One setting worth finding separately. Improve Siri & Dictation, under Settings, Privacy & Security, Analytics & Improvements, controls whether Apple stores audio for human review. It’s opt-in, and Apple says reviewed transcripts may be kept for more than two years. Leave it off if privacy is the point. You can also delete Siri and Dictation history from Siri settings.
How do you get offline voice typing on Android? #
It varies by manufacturer, but on most phones with Gboard:
- Open Gboard settings (long-press the comma, or tap the gear in the keyboard’s toolbar).
- Go to Voice typing.
- Look for an offline speech recognition or language download option, and download your language.
- Test it in airplane mode with Wi-Fi off.
Pixel phones handle more speech processing on the device than most. Samsung phones may use Samsung Keyboard’s voice input or Gboard, so the setting lives in different places. If you can’t find an offline option, your language may not have one yet.
How do you transcribe recordings offline? #
Keyboard dictation is live only: you speak, it types. For voice memos, lectures and meetings you already recorded, you need a transcription app.
Built into iPhone. Voice Memos can show a transcript for recordings, depending on your language and region. Live Captions captions audio playing on your phone, processed on the device. In iOS 26 Apple introduced a newer on-device speech-to-text API for developers called SpeechAnalyzer, so more third-party apps can transcribe without a server.
Built into Android. Google’s Recorder app on Pixel phones transcribes recordings on the device as you record, and the text is searchable afterwards. Live Transcribe gives real-time captions of speech around you, made for people who are deaf or hard of hearing; offline use depends on the phone and language. On supported phones, apps can also use Gemini Nano for on-device transcription.
Whisper-based apps. Whisper is a speech recognition model OpenAI released as open source. Because it’s open, developers can run it directly on phones, and it handles many languages well. Google’s AI Edge Gallery app includes Audio Scribe, which transcribes and translates audio with on-device models.
For a dedicated app, Private Transcribe comes from the same developer as this blog’s app. It runs Whisper through whisper.cpp on the phone, works in airplane mode after a one-time model download, and supports 99+ languages. It offers five model sizes from 78 MB up to 574 MB (Whisper large-v3-turbo), most in English-only and multilingual versions, so you can trade speed for accuracy. It imports recordings and video files (m4a, mp3, wav, aac), and every transcript line carries a timestamp. It’s free with ads; a one-time $7.99 Pro purchase removes the ads, adds on-device translation of any language into English, and unlocks .txt and .srt export.
Before you install any transcription app, check whether it runs on-device. Many upload your audio. How to tell if an AI app is really private walks through the tests.
Comparing your options #
| Option | Platform | Live or file | Offline | Best for |
|---|---|---|---|---|
| Keyboard dictation | iPhone, Android | Live | Often | Typing by voice |
| Voice Memos transcripts | iPhone | Recording | Varies | Quick memos |
| Live Captions / Live Transcribe | iPhone, Android | Live | Yes on many phones | Accessibility |
| Pixel Recorder | Pixel | Recording | Yes | Meetings, lectures |
| Whisper-based apps | iPhone, Android | File or recording | Yes | Long audio, many languages |
| Cloud transcription services | Any | Both | No | Noisy audio, speaker labels |
Why is offline speech recognition harder? #
The models are smaller. Cloud speech services run large models on servers. On-device models have to fit in memory alongside everything else and keep up with your voice. Whisper shows the trade-off directly: it comes in sizes from tiny (about 39 million parameters) to large (about 1.5 billion). The larger versions are noticeably more accurate with accents and background noise, and they need more memory and time.
Noise and crosstalk. Cafés, cars and open offices are where accuracy falls first. Cloud systems generally cope better with messy audio.
Vocabulary. Names, drug names, technical terms and brand names trip up every system. Cloud services sometimes let you add custom vocabulary; offline keyboard dictation usually doesn’t.
Battery and heat. Continuous transcription of a long meeting keeps the chip busy. Short dictation is negligible; an hour of transcription will show up in your battery stats.
How do you get accurate results? #
- Get close to the speaker. Distance and room echo cause most errors.
- Speak in phrases, not word by word. Models use context to pick the right word.
- Say punctuation out loud (“comma,” “period,” “new line”) if your keyboard doesn’t add it automatically.
- Set the language instead of relying on auto-detect, especially for short clips. Dictating Spanish with the keyboard set to English gives garbled text.
- Use a bigger model for important audio. In Whisper apps, a larger model is slower and more accurate.
- Proofread names and numbers. That’s where errors hide and where they cost the most.
Turning speech into notes with an offline AI #
A transcript of a one-hour meeting is a wall of text. A local language model turns it into something useful without sending it anywhere. Copy the transcript into Personal LLM, or save it as a text file and attach it, then ask:
- “Summarize this meeting in five bullet points.”
- “List every action item with the person responsible.”
- “What decisions were made, and what’s still open?”
- “Write a short follow-up email to the attendees.”
- “Pull out every date and deadline mentioned.”
One thing to be clear about: Personal LLM has no voice input and doesn’t record or transcribe audio. It works with the text once you have it, and it can read its answers aloud with your phone’s on-device text-to-speech. Your keyboard’s dictation button also works in its chat box like any other text field, so with on-device dictation you can speak a rambling thought, ask the model to turn it into a checklist, and keep the whole chain on the phone.
For long transcripts, ask about one part at a time or attach the file so the app can find the relevant passages. Our guide to summarizing text offline has more prompt ideas, and how to get better answers from a local AI helps with messy dictated input. Check any summary against the transcript for anything that matters, since summaries drop and blur details; see why AI makes things up.
Frequently asked questions #
Does iPhone dictation work in airplane mode? #
It does if your iPhone processes dictation on-device for the language you’re using. Check the note under Dictation in Settings, General, Keyboard, then test in airplane mode with Wi-Fi off.
Does Apple keep my dictation recordings? #
Not if dictation is processed on your iPhone. For server-processed requests, Apple says it doesn’t retain audio unless you opted into Improve Siri & Dictation, which you can switch off under Privacy & Security, Analytics & Improvements.
Is offline speech recognition as accurate as online? #
In a quiet room with a common language, it’s close. With background noise, strong accents, specialist vocabulary or less common languages, server-based systems are usually more accurate because they run larger models.
How do I transcribe a voice memo offline? #
Use a transcription app that runs its speech model on your phone, such as a Whisper-based app. Check that it works in airplane mode before trusting it with private recordings, since many transcription apps upload audio.
What is Whisper? #
Whisper is an open-source speech recognition model released by OpenAI. Because it’s open, apps can run it directly on phones and computers, offline, in many languages.
Can I talk to an AI chatbot offline? #
Indirectly. Speak into your keyboard’s microphone in a local AI app’s chat box, and with on-device dictation the words never leave the phone. The model answers on the phone, and apps with on-device text-to-speech read the answer back. No wake word, and no hands-free mode.