Install on Android in under a minute.
Scan the QR if you're reading on desktop, or tap the Play badge if you're already on your phone. Both point to the same build.
Yaps é o melhor aplicativo de ditado móvel para Android. Apple Dictation é o melhor integrado no iPhone. Wispr Flow é a escolha de plataforma cruzada mais forte. Se isso é tudo que você queria saber, pode parar por aqui. Se você deseja uma compensação honesta e profunda por trás de cada escolha após algumas semanas de testes diários, esta é a análise.

We spent the last couple of weeks installing mobile dictation apps on real phones and using them for real things. Quick messages, long emails, voice memos on walks, rambling thinking-out-loud sessions, the occasional technical name that trips every speech model up. The ranking below is what we landed on after that, with the surprises and annoyances each app threw at us along the way.
Before we start: Yaps is our app. We built it, we use it every day, and we are obviously biased toward "on-device is the right default for your voice". We still tried to test the others fairly, and you will see places below where another app wins a specific task. Yaps is also Android-only on mobile today. If you are on iPhone, one of the other picks is the right answer for your phone, and we will point you there without hedging.
The thing that kept coming up in testing was not accuracy or speed. Those are close enough across the top apps that most people will not tell them apart in a text message. What varies is what the app has to know about you to work at all. Some want an accessibility service. Some want screen capture. Some upload your audio through partner companies you have never heard of. A couple just ask for a microphone and leave it at that. Mobile dictation has quietly split into two categories along that line, and which one you pick depends on how much of your digital life you are willing to hand over for a good transcription.
Scan the QR if you're reading on desktop, or tap the Play badge if you're already on your phone. Both point to the same build.
If you are on Android, install Yaps. It is the only full AI keyboard on this list that keeps every piece of the speech pipeline on your phone, cleans up filler words and list formatting without a cloud round trip, includes a notepad for longer captures, and asks for exactly one sensitive permission. Setup takes under a minute, and yes, we are biased, but we were also honest about where it falls short further down.
If you are on iPhone, use Apple Dictation (free, built-in, genuinely good on the 2026 models) and add Wispr Flow on top if you want aggressive AI rewriting and are willing to pay for it. Yaps is not on iOS yet, so we will not pretend otherwise.
That is the quick answer. Here is how the full list shakes out, with the platform each app runs on and the one thing you give up with each.
We weighed four things, in order, and the order matters.
Transcription accuracy is not on that list, and that was a deliberate call. In 2026 the best engines (Parakeet on-device, Whisper-class cloud models, Apple's on-device model, Google's speech engine) are close enough that we stopped being able to tell them apart in a text message. What we could tell apart was whether the app was making itself visible in our day. Floating bubbles get in the way. Accessibility-service prompts during setup feel heavy. Cloud uploads make the app slower on bad mobile data. A keyboard that just replaces your keyboard gets out of the way.
Platform: Android 8.0 and later. iPhone: coming, not shipping yet.
Yaps is a full Android keyboard with voice dictation built directly in. Tap the microphone, speak, and the words go into whatever app you are typing into. Every step of the speech-to-text pipeline happens on your phone. Nothing is uploaded, nothing gets transcribed in the cloud, and nothing is saved anywhere outside your device unless you choose to sync it.
The first thing worth saying about Yaps is how much engineering has clearly gone into it. Most new voice apps on Android in 2026 are effectively a UI wrapper around a cloud speech API. Yaps is not that. To run the full pipeline on a phone, they rebuilt the entire keyboard stack from scratch, with a local speech engine, a local cleanup model, a local text-to-speech engine, and a full QWERTY underneath all of it. That is the kind of work well-funded startups in this space usually skip because it is hard and slow. Seeing it actually shipped, and shipped well, is the thing that made us willing to write this post.
The moment the architecture clicks is the first time you dictate a message on a train with no signal and the words just appear. No spinner. No "reconnect and try again". That is what we mean when we say the pipeline runs locally; the phone does the whole job, and that alone gives Yaps the lowest end-to-end latency of anything we tested. There is no upload, no round trip to a remote server, no partner AI provider for the cleanup step. The words just appear. On top of that, Yaps streams the transcription as you speak rather than committing only once you stop, which makes it easier to catch a misheard word mid-sentence and correct it without waiting. Wispr Flow, Typeless, and most of the other serious dictation apps in this category wait until you finish before they show the result. Yaps does not, and the difference in feel is bigger than the millisecond numbers suggest.
We are biased, obviously. We built this. When we set Yaps up next to Wispr Flow and Typeless during testing and gave the same sentence to all three, the two things that stood out were the permission list at setup (Yaps asked for microphone access and not much else) and the lack of a floating bubble covering half the screen.
Parakeet TDT 0.6B via sherpa-onnx. Works offline, in airplane mode, and underground.
Glide typing, emoji panel, clipboard history. Replaces Gboard or SwiftKey as your keyboard.
Ums, uhs, repeated words, and self-corrections get removed. Lists, numbered steps, and punctuation get auto-formatted. No cloud round trip.
A full notepad inside the app. Continuous capture for thinking out loud, segments by pause, searchable locally, no account required.
Highlight text in any app and hear it in a natural Kokoro voice, on-device.
Turn a read-aloud session into a captioned video with word-level timing.
Microphone. That is it. No accessibility service, no screen capture, no draw-over.
No network round trip because it runs on the phone, and the transcription streams as you speak rather than waiting for you to stop. Easier to catch misheard words mid-sentence and correct in flight.
Downloadable voices already ship through the Yaps store. The natural next step is users extending functionality over time without needing a new app release.
Yaps is the strongest voice keyboard on Android today, and none of the gaps below change that. They are just the honest things we would want a prospective user to know before they install.
Learn more: Yaps Android app launch | Wispr Flow vs Yaps on Android | Voice notes feature
Platform: Android 13 and later, iOS 18.3 and later.
Wispr Flow is a genuinely good product. That is the thing that makes what we are about to say uncomfortable. The AI rewriting quality is class-leading. The cross-device sync between Mac, Windows, iOS, and Android is real. The command mode (highlight text, speak a rewrite instruction) is novel and clever. Wispr Flow is a well-engineered piece of software built by a well-funded team, and on raw output it is the strongest AI writing engine in this category.
Where Wispr Flow shines is the rambling email. We tried the exact same three-minute "here is what I need from the team next week, with reasons, with caveats, with a reminder about Thursday's thing" dictation into all five apps. Wispr Flow's cleanup pass turned it into an email we could have sent. The others gave us something that still needed editing.
Then comes the UX, and honestly, this is what surprised us most. The packaging on Android is a separate problem from the pipeline. Wispr Flow runs as a floating bubble that sits on top of your screen. When the keyboard is up, you have the keyboard taking real estate and the floating button sitting on top of everything else at the same time. When the keyboard is down, the bubble stays visible, still competing for taps with whatever app is underneath. There is no state where it gets out of the way. On a small phone it gets aggressive fast, and Play Store reviews reflect it; complaints about the floating button are one of the most consistent strands of negative feedback on the Android build. The core dictation engine is strong. The wrapper around it is the thing you actually touch every time you use the app, and the wrapper is surprisingly weak. In practice, this is probably the biggest everyday issue with Wispr Flow on Android.
Then comes the privacy cost on top. On Android, Wispr Flow asks you to approve five or six different permission sets during setup: accessibility service, draw-over-other-apps, screen capture for the context feature, microphone, and a few smaller ones. Taken individually, each has a plausible reason. Taken together, they add up to essentially complete visibility into your device. What you are typing, what is on your screen, what apps you have open. That is a lot of trust to extend to any third-party app, and more so to a cloud service whose AI providers have, in the past, used user audio for training, with the opt-out buried several screens deep. On iPhone the setup is lighter because iOS simply forbids half of those permissions, but your audio still ends up on Wispr's servers and passes through partner AI providers for the rewriting step.
Strong product, but at what cost? Once you add up the floating-button friction and the permission prompts, the answer is your screen real estate and your privacy. For users who genuinely need the best paragraph-level rewriting on mobile and are explicitly comfortable with both trades, Wispr Flow is the right product. For everyone else, the cost starts to look steep.
Learn more: Wispr Flow vs Yaps on Android | Wispr Flow vs Superwhisper privacy
Platform: iPhone and iPad, built into iOS 26 and later with Apple Intelligence.
If you are on iPhone and have not tried Apple Dictation on the 2026 models, try it first, before anything else on this list. It has quietly got very good. Tap the microphone on the iOS keyboard, speak, and the words appear. On Apple Intelligence devices, dictation runs entirely on-device, auto-punctuation works, and for everyday messages the whole experience is fast enough that paying for an alternative starts to feel unnecessary.
We did a week of testing on an iPhone 17 Pro alongside a Pixel 9, using Apple Dictation on one and Yaps on the other for the same volume of messages. For short-to-medium dictation (text messages, quick emails, Slack replies), the difference in accuracy was a coin flip. The difference in what the apps were doing in the background was not.
The obvious limitation with Apple Dictation is that it is iPhone-only. Android holds most of the global smartphone market, so the best-in-class built-in option on one platform has exactly zero reach on most of the phones in the world. If your household or team mixes Android and iPhone, Apple Dictation does not help you standardise anything. It is a great answer for iPhone users and no answer at all for anyone else.
If you are on iPhone and you want free, private-by-default, "just works" dictation without a subscription, Apple Dictation is the first app to try and the one most people stop at. The step up from there is Wispr Flow on iOS for aggressive AI rewriting. We want to be straight with you: there is no third-party iPhone keyboard today that matches what Yaps does on Android. That is the honest reason we are not crowning an iPhone-specific voice keyboard at number one. When an iPhone build of Yaps lands, this ranking will change. Until then, Apple Dictation is the right place to start.
Learn more: What data does Apple Dictation actually send? | macOS Tahoe dictation review
Platform: Android. Not available on iPhone.
Typeless is also a genuinely strong product. A voice-first keyboard. AI cleanup that strips filler words, self-corrections, and repetition. Auto-formatting for lists. A personal dictionary for jargon. A command mode for editing by voice. On the Play Store screenshots it looks like a direct competitor to Yaps, and for the happy-path workflow it is.
The friction we hit was not the recognition quality. It was what happens the moment recognition gets something wrong. The first time Typeless misheard a name in testing, we went to tap out a correction and realised there was no keyboard to tap on. Typeless is a microphone button with voice commands, not a QWERTY with voice on top. When a mistake lands, you have two options: redictate the sentence from scratch, or switch back to your previous keyboard through the Android input picker to patch the error. Both are noticeably more effort than just tapping the letter you wanted. After the third or fourth misheard name, that friction compounds.
The command features ("make this more professional", "bullet this", "translate to French") are a genuinely interesting direction, but command mode is not how most people use a keyboard most of the time. The core usage pattern is type a thing, see a thing, adjust a letter, move on. That loop is where Typeless struggles.
Pricing is the other thing to flag. The free tier covers 8,000 words per week, which looks generous, but the paid tier is where the cost gets steep. Pro on the monthly plan is $30 per month (roughly £27 in the UK, which is closer to $36.50 once currency and regional markup land). The annual path works out cheaper at $144 per year, or effectively $12 per month amortised, but you have to commit up front. Compared to Yaps' freemium model on Google Play and the free options already on your phone, the monthly price specifically is one of the most aggressive in the category.
Typeless is a fair pick if you are comfortable dictating almost everything and happy to use voice commands for edits. If you want a voice-first keyboard that still hands you a full QWERTY the moment you need to fix a misheard word, Yaps is the app that keeps both modes under the same roof.
Platform: Android 6.0 and later, pre-installed on most phones.
Gboard is a genuinely great typing keyboard. It is a staple on Android for good reason: glide typing, emoji, mature layouts, a long track record. Nothing on this list is trying to argue otherwise. The problem is that Gboard was built as a typing keyboard first and a dictation tool second, and the dictation side shows its limits the moment you lean on it.
We dictated the same rambling voice note into Gboard and then into Yaps, and the Gboard version came back as a single unbroken paragraph of half-formed thoughts with every um and uh still in. Gboard gives you whatever the speech model heard, verbatim, and leaves the cleanup to you. There is no filler-word removal, no list formatting, no self-correction handling, nothing that resembles the kind of output a more modern voice keyboard produces out of the box. The dictation feature itself is also more limited than it looks: getting on-device recognition reliably active takes a few extra steps (download the language pack, confirm settings, sometimes wait for a model sync), and even then the output feels like a raw transcript rather than something ready to send.
None of that makes Gboard a bad keyboard. It just means the dictation layer is the weakest part of Gboard, not the core of it. If you already use Gboard and you occasionally dictate, you will get by. If dictation is what you actually want, this is not where it lives.
Gboard is the right answer if you do not want to think about voice typing enough to install something else. It is rarely the best answer once you do.
Learn more: Voice data privacy in 2026 | Offline dictation guide
Here is the scoreboard from our testing. All five apps, across the dimensions we actually cared about. Scroll sideways on mobile.
| Capability | Yaps | Wispr Flow | Apple Dictation | Typeless | Gboard |
|---|---|---|---|---|---|
| Platform | Android | Android + iPhone | iPhone only | Android | Android |
| Input surface | Voice + QWERTY | Overlay / keyboard | Mic in system keyboard | Voice-only | Type + voice |
| Fallback QWERTY typing | Yes, in-app | Via host keyboard | Yes (iOS keyboard) | No | Yes |
| Speech processing | On-device | Cloud | On-device (AI devices) | Cloud | Hybrid |
| Works offline | Yes | No | Partial | No | Partial |
| Filler word removal | Yes, on-device | Yes, cloud | No | Yes, cloud | No |
| Auto-formatted lists | Yes | Yes | No | Yes | No |
| Built-in notepad | Yes | No | No | No | No |
| Accessibility service | Not used | Required on Android | Not used | Not used | Not used |
| Sensitive permissions | 1 | 5-6 (Android) | System-level | 1-2 | 1-2 |
| Trains on your data | Never | Opt-out | Opt-in only | Policy-based | Via Google account |
| Minimum OS version | Android 8.0 | Android 13 / iOS 18.3 | iOS 26 for on-device | Android 8.0 | Android 6.0 |
| Price | Freemium | $15/mo or $144/yr | Free | $30/mo or $144/yr | Free |
Cloud pipeline
Typeless and Wispr Flow both upload audio to remote servers on both Android and iPhone. Gboard falls into cloud mode on older devices. Privacy depends on a vendor's policy, its partners, and any future policy changes it decides to ship.
On-device pipeline
Yaps on Android and Apple Dictation on Apple Intelligence iPhones both transcribe locally. No audio pipeline leaves the device for a policy change, server breach, or subpoena to expose. The audio is not collected, so it cannot be used for anything else.
The question most round-ups skip is the one we came back to over and over in testing. Where does your voice actually go?
Yaps on Android and Apple Dictation on Apple Intelligence iPhones both process audio on your phone. Typeless and Wispr Flow upload audio to servers on both platforms. Gboard is a split case: the on-device engine stays local on newer Pixel and Samsung devices, the cloud engine does not, and which path your dictation takes depends on your device, your settings, and whether you have bothered to install the local language pack.
Privacy by architecture, not by policy. If the audio never leaves the device, no policy change or breach can expose what is not there.
Yaps for Android
For most people the difference between privacy-as-policy and privacy-as-architecture is abstract. For anyone in healthcare, law, or any regulated industry, it is the entire reason to pick one app over another.
Start with the phone in your pocket, then pick the workflow that describes you.
For most Android users, the right answer is to install Yaps and stop there. For most iPhone users, the right answer is to keep Apple Dictation, add Wispr Flow only if you write long emails by voice, and keep half an eye on the category because it is moving fast.
Mobile dictation used to be a commodity category. You picked a keyboard, tapped the microphone, and the text came out mostly right. The cost of the feature sat quietly in the background and nobody had to look at it. In 2026 the cost is still there, it just moved: permissions you grant during setup, audio uploads you did not realise you were agreeing to, accessibility services that read what is on your screen, data ending up in training sets you will never see. The apps asking for the most in exchange for turning your voice into text are also the ones with the loudest marketing budgets, which is part of why we wrote this piece.
The market has started to answer. Yaps is our answer on Android, and the one we think most Android users should start with. Apple Dictation has quietly become a solid default on iPhone, and Apple Intelligence is doing more of the work on-device with each release. Wispr Flow is a genuinely useful answer if AI rewriting matters more to you than privacy architecture. The worst options, the ones demanding six permissions and uploading everything by default, are losing ground.
Pick the mobile dictation app that fits your phone, your workflow, and your tolerance for handing audio to someone else's servers. Read the permission list before you tap install. If you are on Android and have not installed Yaps, that is the one sentence of this review we would want you to take with you.
Free tier, one microphone permission, works offline. If it is not for you, Android switches keyboards in one tap. No lock-in, no account required.
Here are the questions people kept asking us while we were writing this.
Yaps, if you are on Android. It is the only full AI voice keyboard on mobile that keeps the whole pipeline on your phone, cleans up filler words and list formatting on-device, ships a built-in notepad, and asks for one sensitive permission. If you are on iPhone, the best mobile dictation app is Apple Dictation for free use, or Wispr Flow if you want AI rewriting and do not mind a subscription. There is no current third-party iPhone keyboard that fully matches what Yaps does on Android.
On Android, Gboard is free and pre-installed, and Yaps has a free tier (1,000 words per week of on-device dictation). On iPhone, Apple Dictation is free, built-in, and genuinely capable on Apple Intelligence devices.
Apple Dictation for most users. It is free, built-in, on-device on Apple Intelligence devices, and private by default. Wispr Flow on iOS is the step up for AI rewriting and cross-device sync, but it is a subscription service and your audio goes to the cloud.
Yaps. Full AI voice keyboard, on-device processing, filler-word removal, list auto-formatting, built-in notepad, one sensitive permission, and works in airplane mode.
Not yet. Android is the only Yaps platform publicly shipping today. The macOS desktop app, the Windows desktop app, and an iPhone build are all in active development, not yet available to install. iOS in particular is a harder lift because of how Apple limits third-party keyboards. When the iOS build is ready we will add it to this ranking.
On devices that support Apple Intelligence, yes. On older iPhones, some dictation still falls back to server processing. Check Settings > Keyboards on your device for the indicator that dictation is being processed locally.
No. Wispr Flow's transcription and AI rewriting both run in the cloud on Android and iPhone. Without a data connection, the app does not work.
Partially. On recent Pixel and Samsung devices with the on-device voice model installed, dictation runs locally. On older devices or certain settings, voice data still uploads to Google's servers. Check your Google account's voice and audio activity settings to confirm what is being kept.
They are marketed around the same idea, but the shape of the two apps is different in two meaningful ways. First, Typeless runs its speech recognition and cleanup in the cloud; Yaps does both on your phone. Second, Typeless is a voice-only interface with voice commands around it, while Yaps is a full QWERTY keyboard with voice dictation built in. That second difference shows up the first time the speech model mishears a name or a technical term. With Typeless you either re-dictate or issue a voice command. With Yaps you just tap the letter you want to change. Yaps also ships a built-in notepad and a system-wide read-aloud feature that Typeless does not.
Yaps on Android asks for one sensitive permission, microphone access, and that is the entire surface. Apple Dictation on iPhone uses system-level permissions rather than app-specific ones. Typeless and Gboard sit in the one-to-two range. Wispr Flow on Android needs five to six, including an accessibility service and screen capture.
Yaps does not. Audio never leaves the device, so there is no training pipeline that could include it. Apple Dictation does not retain audio unless you opt into the "Improve Siri and Dictation" program. Typeless advertises a zero-retention policy but the audio still passes through cloud infrastructure. Gboard and Wispr Flow have varying training-on-user-data policies with different degrees of opt-out. Read each app's settings screen and privacy policy before you install.
Dragon Anywhere is professional dictation software aimed at very specific workflows (medical, legal, long-form transcription) and is priced accordingly. SwiftKey is a Microsoft keyboard with voice typing as a secondary feature, heavier data flows, and no serious AI rewriting. Both have their audience; neither made our top five mobile picks for 2026. See our Dragon Medical alternative post for more on professional dictation.
Yes. Android lets you switch keyboards from the input picker in a single tap, and iOS lets you add third-party keyboards alongside the system keyboard. Install Yaps and Wispr Flow on Android, or Apple Dictation plus Wispr Flow on iPhone, and compare for a week. Most users end up with one primary voice workflow and disable the others.
Yaps, practically speaking. Two reasons. First, it runs locally, so there is no upload or network round trip in the path between you speaking and the words appearing; Apple Dictation on Apple Intelligence devices and Gboard in offline mode share that advantage over the cloud options. Second, Yaps streams the transcription as you speak rather than waiting for you to stop. That makes the experience feel faster than end-to-end numbers alone suggest, because you are seeing text accumulate in real time instead of sitting with a progress indicator. Typeless and Wispr Flow both upload audio and commit only after you finish, which is noticeable even on good mobile data.
Yes, for most use cases. The best on-device models on both platforms now clear 95 percent accuracy for clear speech in a quiet room. Noisy environments, accents outside the training distribution, and technical jargon still cause errors. The gap between the best on-device engine and the best cloud engine is small enough in daily use that most people should pick on privacy, workflow, and cost rather than chasing a few extra percentage points of accuracy.