Voice Typing in Telegram, WhatsApp, Slack, and Discord

I have six messengers myself. Telegram (personal and work), WhatsApp (relatives), Slack (current client), Discord (two community servers), Signal (sensitive conversations), and iMessage via iCloud Web. That's actually why I built Commander Flow the way I did — I no longer wanted to context-switch between messengers by hand.

Each one has its own conventions, pace, and tone. I used to spend mental effort switching between them. Now, by my own design, I have one hotkey and one habit — and I know this carries over to anyone who uses the product seriously.

What my chat day looks like

Mom messages me on Telegram about the cabin. I hold the key, say: "thanks mom, we'll come on Saturday, bringing the meat for grilling, love you." Voice command: "friendly tone, like I'd write to my wife." In the chat: "Thanks, Mom! We'll come Saturday. Bringing the meat for the grill. Love you." Nobody adjusts for the fact I dictated it.

A few minutes later — Slack, client about a deadline. Same key, voice: "business tone, brief." Slack gets: "Got it. Wed at the latest. I'll send a draft for review Wednesday morning." No "well" or "you know." No filler.

And a minute after that — Discord, the gamedev folks discussing a shader. Same key, but now I say: "leave it as I said, no edits." The chat shows: "bro this is straight up fire how you wrote that compute shader." Exactly how I exhaled it.

"Three different tones in three chats back to back — no layout switching, no mode change, no separate app per tone."

Test: "guess which chat I just wrote in"

I once asked my wife, as an experiment, to guess from my messages who I was writing to. She got every one. It wasn't magic — Commander Flow adjusts capitalization, length, emoji use, and formality based on the voice command I gave it.

With Mom — short sentences, exclamation marks, warm phrasing.
With the client — no emotion, just facts, clear timing, no "well actually."
With friends — words from spoken speech, not "corrected" into book English.

Why this matters especially for non-native speakers

Let me tell you about my aunt — she immigrated to Canada and works in HR there. She writes in English at work, in her native language at home, in Ukrainian to old friends. She used to keep a cloud translator in one tab and a cloud style editor in another, all the time.

I gave her Commander Flow for her 60th birthday. She dictates in her native language, says "in business English," and writes to colleagues like a native. Switches the command to "friendly, in Ukrainian" — and writes a warm note to a friend. No layout switching. No tabs. No "what's the best way to say this in English."

At 60, she uses voice input better than many twenty-somethings. For me as the author of this thing, that's an important signal: I'm not building "tech for tech's sake," I'm building a solution to a real language load — and it works for people who are far outside the IT world.

How this makes "writing in chat" cheap

Within a couple of months, my domestic habit changed: I started replying to everyone. I used to ignore messages because "a long reply is too much hassle." Now I open the chat, say one phrase the way I think, and get a clean text in a couple of seconds.

A new rule appeared in my notebook: "if a reply would take less than 5 seconds, reply now." That rule used to be unrealistic, because any reply took at least 30 seconds. It works now.

The result: I've gotten a lot more social. My relationship with Mom improved, I regularly answer old classmates I used to brush off with "let's talk when I have time." I didn't expect this side effect from a piece of technology.

Where I run into friction — and what I'm doing about it as the author

Discord sometimes "eats" the start of a voice command. If Discord is open and an active voice channel is running, there can be a microphone conflict. Fixed by setting Discord to push-to-talk in its mic settings. One-time setup. I keep this in the FAQ — but the honest fix is to detect this situation automatically, and that's on my list.

WhatsApp Web sometimes won't allow text insertion because of its JS interception. Not Commander Flow's problem — it's WhatsApp. Workaround: switch the insert strategy to clipboard-paste for that process. Settings have a per-app allowlist; I extend it regularly based on user feedback.

If you dictate too quietly in a noisy room and ask for "friendly," the model sometimes "imagines" emotion. You speak neutrally and get something overly upbeat. That's about user articulation, not the tool — but I'm thinking about how to soften it on the polishing side.

What I didn't expect from my own technology

The biggest value of Commander Flow in chats is removing the psychological cost of communication. When sending a message is almost free, you become reachable many times more often. That changes your social map.

I've gotten more attentive to people. Not because I changed inside, but because the few-minute gap between "intend to reply" and "actually send" disappeared — the gap I used to lose messages in. I didn't put this side effect in the roadmap — but it's exactly the reason I keep building.

Good tools, it turns out, work exactly like that.