Three different purchases wearing one name
If you only want to send a message by speaking, start with the microphone already on your keyboard. If you want every recording turned into searchable meeting notes, use a transcription service. If you want to dictate privately without a connection, choose an app that runs the recognition model on the phone. Those are three different purchases, even though search results call all of them voice to text apps.
| App | Best for | Where the text lands | Connection |
|---|---|---|---|
| Gboard | The easiest place to start | Any text field | Depends on device and mode |
| LocalType | A complete private voice typing keyboard | Any text field | Offline after model download |
| FUTO Voice Input | Keeping your present keyboard | Any text field through voice input | On-device |
| WhisperIME | Open-source users who accept setup | Any text field through an IME or voice service | Offline after model download |
| Speechnotes | Long notes in a dedicated notepad | Inside Speechnotes | Uses Android or browser recognition |
| Otter | Meetings and recorded conversations | A saved meeting transcript | Online service |
| Live Transcribe | Reading speech and surrounding sounds live | On its own transcription screen | Online or selected offline languages |
Product behaviour above was checked on 21 August 2026. There are no accuracy percentages in that table because recognition quality moves with the phone, the language, the microphone, the room and the model, and a number copied from a product page would tell you less than ten minutes with your own voice.
What counts as a voice typing app
The result is what separates them. Voice typing puts words into the text field you are using now. A voice-notes app gives you a note inside that app. Meeting transcription keeps a recording and produces a document afterwards. Live captioning shows what people are saying so you can read the conversation.
Only the first of those replaces typing across Android. The others can be excellent at their own work, but copying a paragraph out of a notes app every time you answer a message is not a keyboard workflow. This is why lists of the best voice to text apps often feel incoherent: they compare products that happen to use speech recognition and stop there.
The same mixture turns up whichever phrasing you search for. One result is a keyboard microphone, the next a blank page for long-form dictation, the next a recorder that produces text afterwards. Before comparing accuracy or price, remove every option that sends the words to the wrong place.
Four questions did the sorting. Does it write in any app or only its own? Does speech stay on the handset? Can it work without a connection? Is it meant for one nearby speaker, or for a room and a recording? There is no winner across all four because the jobs conflict.
Gboard: best place to start before installing anything
Gboard is the sensible first test because it is already present on many Android phones and its microphone writes into most places where the keyboard opens. Google documents spoken punctuation, word replacement and language switching. On supported Pixel phones, Advanced voice typing adds automatic punctuation, voice editing and the ability to keep typing while the microphone remains active.
The privacy and offline answer is not one sentence for every Android phone. Google says ordinary talk-to-text sends voice to Google for transcription. Its current Advanced voice typing documentation says spoken text stays on supported Pixel devices, with exceptions for features such as Fix it and detailed edits. Availability also depends on the handset and language. That makes Gboard easy and capable, but it means you must check the mode your own phone is actually using.
Choose Gboard if it already works well for your language and you do not need a single offline guarantee across different Android devices. A good built-in answer is still a good answer.
LocalType: best complete offline voice typing keyboard
LocalType is a keyboard rather than a recorder or a voice-notes destination. Tap its microphone while writing a message, email, note, search or form and the result appears there. The same app handles ordinary typing, autocorrect, next-word suggestions, layouts and emoji, so dictation does not sit beside a different keyboard as a separate component.
Speech recognition runs from a model in private app storage on the Android device. Internet access has one purpose: downloading the model you choose. After that there is no cloud transcription route, no account, no advertising, no behavioural tracking and no archive of recordings. App data is excluded from cloud backup, and the microphone is unavailable in a password field.
A single multilingual model covers eighteen languages and can identify the one being spoken. Three sizes, 60, 190 and 539 MB, let an older handset favour speed and a newer one favour accuracy. The middle model turned 6.9 seconds of speech into text in 4.1 seconds on the 4 GB test phone. LocalType runs on Android 8.0 and newer.
There are real costs. LocalType is Android-only, it does not transcribe existing recordings or meetings, and its source is not published. Voice typing is paid after the free trial. If you require inspectable code, FUTO or WhisperIME below is the better place to begin.
Choose LocalType if you want one finished keyboard for typing and dictation, use several languages, and want offline processing to be the architecture rather than a mode you need to check.
FUTO Voice Input: best if you want to keep your keyboard
FUTO Voice Input solves a different problem. It supplies on-device speech recognition as a voice-input component, so you can keep a keyboard you already like. FUTO describes microphone data as always processed on the device and never transmitted elsewhere. The source is published, and the app is available through Google Play, F-Droid and a standalone download.
That separation is its advantage. Changing your speech input does not require changing your layout, autocorrect or typing habits. It is also the extra moving part: Android and the chosen keyboard have to hand the microphone action to the voice-input service correctly. FUTO also offers its own fully offline keyboard if you prefer an integrated product.
It suits the reader who cares about open source and local processing but will not give up the keyboard under their thumbs.
WhisperIME: best for an open-source setup you control
WhisperIME is an open-source Android input method built around Whisper. It can run as its own input method, work with a compatible keyboard microphone button, or be selected as a system voice-recognition service. Its model download is roughly 435 MB. Recognition then runs offline, with a compact English model and a slower multilingual option.
The project documents a thirty-second limit for each recording and, on phones that refuse to expose third-party voice services, a setup route involving Android debugging commands. Neither is fatal for somebody who deliberately chose a small open-source project. Both matter to somebody expecting a polished Play Store installation.
Free, inspectable and offline, provided you are willing to treat the installation as a small technical project.
Speechnotes: best for a long dictated note
Speechnotes gives you a dedicated notepad on Android. That is useful when the destination really is one long note: a draft, a train of thought, an outline or text you plan to edit before moving elsewhere. Its wider service also handles uploaded audio and video, but that is transcription rather than phone-wide voice typing.
The company says dictation in its Android app is delegated to the operating system, while its separate transcription service sends recordings to speech engines. In other words, installing Speechnotes does not by itself tell you where Android performs live recognition. What it adds is the clean notepad and a workflow built for sustained dictation.
Sit inside one document and speak for a while and it earns its place. Reach for it to dictate twelve words into whichever app is already open and it will annoy you.
Otter: best for meetings, not everyday typing
Otter records and transcribes in-person and virtual meetings. Its Android app belongs to a meeting system that can preserve the conversation, make it searchable and connect it with online meeting services. That is valuable work which a keyboard is not designed to do.
It is also why Otter should not win a list about replying to messages by voice. A meeting tool expects a session, a saved transcript and often several speakers. Voice typing expects one speaker close to the phone and a result inserted into the current field. Choosing between them is choosing the output, not the recogniser.
When the audio itself matters afterwards, Otter is the right answer. When the finished text is all you wanted, a keyboard is.
Live Transcribe: best for reading a conversation
Google Live Transcribe listens to speech and surrounding sounds and displays them on its own screen. It can label sounds, search a transcription, use an external microphone and notify you when a chosen name is spoken. On Pixel and selected Android devices, supported languages can be downloaded for offline transcription.
That makes it an accessibility and live-captioning choice, not a way to fill every text box on the phone. If you need to follow what another person is saying, Live Transcribe is closer to the right tool than any keyboard in this list. If you need to write an email, it leaves you with an extra transfer step.
Its job is reading speech as it happens. It does not pretend to be a writing keyboard and should not be judged as one.
How to choose the best voice to text app for you
- 1 Name the destination. Any text field means a keyboard or voice-input service. One long note means a dictation notepad. A reusable recording means transcription. A readable conversation means live captioning. 2. Run the airplane-mode test. Finish any model download, turn off Wi-Fi and mobile data, then dictate. The result is more useful than an ambiguous claim that notes are stored offline. 3. Test your real language. English reviews do not predict names, accents and punctuation in another language. Use the sentences you actually send and include the proper nouns that usually fail. 4. Count the taps. A highly accurate transcript is still the wrong tool if every short reply requires opening another app, starting a note, stopping, copying, switching back and pasting. 5. Check the honest disadvantage. Open source can mean more setup. On-device models occupy storage and may run slower on old hardware. Cloud services need connectivity and receive audio. A product that hides its trade-off has not removed it.
The shortlist, and where it stops
For ordinary Android writing, the shortlist is Gboard, FUTO Voice Input and LocalType. Try Gboard first if convenience is the whole question. Choose FUTO when you want inspectable on-device voice input behind a keyboard you already trust. Choose LocalType when you want the keyboard and offline dictation designed as one product, especially across several languages.
For everything else, use the specialist. Speechnotes is more comfortable for a long note. Otter is built around meetings and retained transcripts. Live Transcribe is made for reading speech in the room. WhisperIME is the appealing answer when control and open source matter more than a guided setup.
LocalType publishes this comparison and is not affiliated with Gboard, FUTO, WhisperIME, Speechnotes, Otter or Live Transcribe. Which one wins depends on where you need the words to land, and that is a question about your own day rather than about recognition quality.