Braina's phone app is a remote control
Braina is a Windows assistant with dictation as one of its abilities. It can transcribe using online services or using language models running on the PC, it covers a very wide set of languages, and it does a good deal more besides dictation: voice commands, chat, controlling the machine by speaking to it.
There are Braina apps on Android and iOS, and this is the part that trips people up: they exist to operate the PC remotely. The dictation still belongs to the computer, which has to be switched on, awake, reachable and running the software before the phone in your hand produces a single word.
Someone typing "Braina for Android" into a search box is therefore usually after one of two very different things.
A remote control for the PC. In that case the existing mobile app is the answer and nothing here replaces it. Sitting on the sofa, telling the desktop in the other room to do something, is a real use and it works.
Dictation on the phone, the way Braina does it on the desktop. Nothing in the mobile app covers that one, because the recognition never moves off the computer. The phone is a microphone and a screen.
The second reading is where LocalType comes in. If what appealed to you was dictation handled by models running on hardware you own, with no service in the loop, the phone version of that idea is a keyboard with the model inside it.
LocalType does the dictating on the phone. It is an Android keyboard with a microphone key, and the speech model that recognises you is a file on the handset, with no PC in the picture and no service behind it. It runs on Android 8.0 and newer, and it arrives through Google Play.
LocalType is a separate product, not affiliated with Braina or Brainasoft.
Why the desktop version cannot simply move over
Braina's ability to run language models on the machine is genuinely appealing, and it is easy to assume that capability travels to a phone. Two things stop it, and only the first is about hardware.
A desktop has memory to spare, active cooling, and often a graphics card willing to do the heavy lifting. A phone has none of those. Instead there are a few gigabytes of memory shared with everything else, no fan, and a battery it is trying not to drain. Anything running there has to be small enough to load quickly, finish before you lose patience, and stop cleanly the moment you stop talking.
That constraint is the entire design problem LocalType exists to solve, and it is why the app ships more than one model. The smallest is 60 MB, aimed squarely at handsets that are short of memory. The middle one is 190 MB and is the default recommendation. The largest is 539 MB and reads accents and noisy rooms best, while asking you to wait: measured on a phone with 4 GB of memory, the middle model turned 6.9 seconds of speech into text in 4.1 seconds, and the largest takes roughly double the time you spent speaking.
One is kept at a time, so moving between them replaces the file rather than accumulating them, and going back down a size later costs you nothing except another download.
The second obstacle is the shape of the writing itself, and it explains why desktop dictation software and phone dictation software end up looking so different.
At a desk you dictate in long stretches. You are alone, or near enough, you have a document open, and you talk for a paragraph at a time before stopping to read it back. Software built for that can afford a setup ritual and a window of its own.
On a phone almost nothing works that way. You dictate one sentence into a reply while walking. You add a line to a note in a shop. You answer a message with someone sitting next to you, so you say it quietly and quickly. These are five-second interactions, and any friction at all, an app to open, a permission to confirm, a result to copy, means you give up and type instead.
So LocalType puts the microphone on the keyboard itself, and not into an app of its own. Not because a separate app could not transcribe just as well, but because in a five-second window nobody opens one.
No PC required, and no connection either
Because the model sits on the phone, dictation does not depend on a desktop being awake or a network being available. On a train, on a flight, abroad with data off, in a basement: the app behaves identically, because nothing was travelling in the first place.
The app goes to the network for exactly one purpose, collecting the speech model you chose, and afterwards it has no reason to go back. Nothing in it sends audio anywhere. No account exists, nothing is tracked inside the product, and the app's data stays out of Android's cloud backup on purpose.
The rest follows from having no server behind it. The model file sits in the app's own private storage, where no other app can read it. No archive of your recordings is kept, because the audio becomes text and that is the end of its life. There is no advertising in the product, so nothing you dictate is feeding a profile. And in password fields the microphone switches itself off, which is the one place where an accidental dictation is genuinely expensive.
Words you add by hand, a colleague's surname, a product name, the acronym your department cannot stop using, stay on the handset as well, and go to the speech model as context so they come out spelled the way you spell them.
Eighteen languages, chosen from the recording
One multilingual model in LocalType covers eighteen languages. It works out which of them you are speaking from the audio itself, so moving between languages between messages requires nothing from you. Set a specific language and the layout, the dictionary and the dictation language line up together.
That is fewer languages than a desktop tool with server-class hardware behind it can offer, and the reason is the hardware constraint: a model that fits on a phone cannot cover everything equally. Accuracy across the eighteen is also uneven, because training material is.
If your working language is one of the ninety Braina lists and not one of our eighteen, this comparison ends there, and the languages page is the fastest way to find that out.
An assistant, versus a keyboard
The other real difference is scope, and it cuts against us as much as for us.
Braina does many things. It answers, it searches, it automates, it controls the machine. Dictation is one capability inside a broader assistant, running on a desk machine with the room to host all of it, with a phone app to reach it from across the house.
LocalType does one thing. It is a keyboard. There is no assistant, no chatbot, nothing that answers questions or writes text on your behalf. You speak, it writes down what you said, and that is the whole feature. When it is not listening it is an ordinary keyboard with autocorrect, next-word suggestions, a dictionary per language, the standard layouts including QWERTZ and AZERTY where those are expected, your own added vocabulary, and a full emoji set.
Whether narrower is better depends entirely on what you were looking for. If you wanted a voice assistant, this is not one, and it will not become one. If you wanted the dictation part without everything attached to it, on the device that has the least room and gets used the most, that is precisely the offer.