Voice note or text, and who pays for it
When typing feels like too much effort in a chat, most people reach for the record button. It is the obvious move: you speak, it sends, you are done.
The cost lands on the recipient. A voice note has to be listened to, at your pace, in a place where audio is acceptable, in full, with no way to skim, and with nothing searchable left behind afterwards. A forty-second message that contains one useful sentence still takes forty seconds.
Dictation gets you the same effort saving and produces text. The reader glances at it. They find it again next month by searching a word in it. They read it in a meeting without headphones.
For a lot of everyday exchanges, that is simply a better outcome for everybody, and it costs you nothing extra.
Where voice notes still win
Sometimes a voice note is the right answer, and for real reasons.
Tone carries in a voice that does not survive text. An apology, bad news, a joke that would read as sarcasm. Complicated explanations where you would be gesturing if you were there. Anything emotional between people who know each other well.
And situations where you genuinely cannot look at the screen at all, which dictation still requires briefly.
The rule of thumb most people converge on: information goes as text, feeling goes as audio.
What makes chat dictation different from other writing
Messages are short, frequent and interleaved with everything else you are doing, which changes what matters in a dictation tool.
Starting has to be instant. If dictating requires opening an app, waiting for it to load, speaking, copying and pasting, you will type instead every time. For a twelve-word reply that overhead is longer than the typing it replaced.
It has to work in whichever app the conversation is in. Conversations are scattered across several messaging apps, plus email, plus whatever else. A tool that only works in one of them is a tool you will forget about.
Corrections have to be easy. Chat is unforgiving of small errors because messages are short, so one wrong word is a larger share of the message.
That combination is why a keyboard suits messaging better than a dictation app does. The microphone is already there when the conversation opens, the words go straight into the box you were about to type in, and you can fix a word by tapping it like any other typo.
The habits that make it work
Say it as you would speak it. Chat is informal, and dictation is naturally informal, so this is the one kind of writing where the register matches without any effort. Do not construct careful sentences.
One thought per message. Speaking encourages long unbroken paragraphs, which read badly in a chat bubble. Say a sentence, send it, say the next one. That also matches how conversation actually flows.
Glance before sending. Names and unusual words are where errors cluster, and a chat message goes out instantly with no undo in most apps.
Do not dictate the punctuation. Modern systems infer it from your rhythm. Saying "comma" out loud usually produces the literal word.
Group chats and the awkward part
One honest limitation of dictating in company: you are saying the message out loud.
In a shared office, on a train, in a room with the person you are messaging about, that is a problem no software solves. Dictation is quietly excellent when you are alone and awkward when you are not.
What helps is that a keyboard makes the switch invisible. You dictate the parts that are harmless, type the part that is not, and neither requires leaving the conversation. Dictation living in a separate app, with its own screen, cannot do any of that.
Reading the room before you speak
The social side decides whether you keep using dictation at all.
Dictating in front of other people is a judgement call, not a technical one. The useful distinction is between what you would be happy saying aloud in that room anyway and what you would not. A grocery reminder, a time, a confirmation: fine. Anything about the person sitting opposite: obviously not.
The practical trick is length. Short dictations attract no attention because they sound like half of a phone call. Long ones make the whole carriage aware of your Thursday.
None of that is a software feature. It is just what the promises about dictation leave out when they tell you it will change how you use your phone, and it is why a keyboard that lets you switch between speaking and typing without leaving the message ends up being the version people stick with.
The microphone is already in the conversation
LocalType is a keyboard, so the microphone key is already there when the conversation opens, in every app that accepts text: messaging, email, notes, browsers, forms. Nothing is copied, the clipboard is untouched, and you can move between typing and speaking inside the same sentence.
The speech model runs on the phone, and that matters more in messaging than anywhere else, for two practical reasons.
The first is that messages get written in exactly the places where a connection fails. Trains, lifts, basements, aeroplanes, abroad with data off. Dictation that depends on a server stops there, and a keyboard that stops working is worse than useless. LocalType behaves the same in airplane mode as at home, because after the model download nothing asks for the network at all.
The second is what messaging contains. Personal messages are the most private writing most people do, and with on-device recognition there is no upload, no account attached to the keyboard, no advertising, no behavioural tracking inside the product and no archive of your recordings anywhere. The model sits in storage only the app can read, and the app's data is deliberately excluded from Android's cloud backup.
Once the message is sent, it belongs to the messaging app and its rules. The keyboard's part ends when the text appears in the box.
Three model sizes, at 60, 190 and 539 MB, let you trade speed against accuracy. For chat the smaller and middle options usually feel best, because replies are short and the wait is the thing you notice most. Eighteen languages are handled by one multilingual model, and it works out which you are speaking per recording. Group chats that switch language halfway benefit from that more than anything.
In password fields the microphone switches itself off, which matters on a phone where a login screen is one tap from a conversation.
The net effect, once it becomes habit, is that the messages you would have left unanswered until this evening get answered now, in text, in about the time it takes to read them. That is the whole benefit, and it is a larger one than it sounds, because the reply you send today is worth considerably more than the better-worded one you send on Thursday.