How to dictate messages in Microsoft Teams and Discord

Picture a stadium built entirely for your voice. Seventy thousand people roar on the terraces, the commentators never stop talking, the sound system is engineered to carry a whisper to the back row. Then you walk over to the ticket counter to leave a note, and there is a small sign taped to the glass: no talking here, please write it down. That is exactly what Microsoft Teams and Discord feel like the moment you try to dictate a chat message.

Both apps are obsessed with your voice. Teams runs your meetings, your calls, your live captions. Discord is wall-to-wall voice channels where people talk for hours. Yet the one narrow place where you actually type, the message box, is the ticket counter with the no-talking sign. Neither app turns your spoken words into a typed chat message. This guide explains why that gap exists, then gives you a real, free way to talk into those boxes anyway, with the rough edges stated honestly so you know what you are getting.

The voice is all on the field, none at the counter

Here is the confusing part, and it is worth being precise about. Teams and Discord are loud with speech features, but every one of them lives on the field, not at the counter. Teams has live captions and meeting transcription. Those write down what people say in a call. They do not put your spoken words into the compose field where you type a chat. Teams also has a separate voice-message feature in some clients, but that records and sends an audio clip. It does not turn your speech into editable text you can fix before sending.

Discord is the same story from a different angle. It has voice channels where you talk on a call, optional transcription for voice-channel audio you receive, and a /tts command. It is tempting to think /tts is dictation, but it is the exact opposite: text-to-speech reads a message you already typed out loud in a robotic voice. It does not listen to you and type. Discord has zero native speech-to-text into the message box. You can confirm the direction of /tts yourself in Discord's Text-to-Speech 101.

So to be plain about it: there is no microphone button in the Teams or Discord compose bar that types what you say. If you want to dictate a message, you fall back to the layer underneath both apps, the operating system itself.

The free workaround: dictate with your operating system

Windows and macOS both ship free dictation that types system-wide. It is not glamorous, but it is genuinely useful and it costs nothing. Click into the Teams or Discord message box first so the cursor is blinking there, then start dictating. Whatever you say lands in that box.

1. 1

   On Windows: press Windows key + H

   With your cursor in the message box, press Win + H. A small floating Voice Typing toolbar appears and starts listening. Speak, and the words type into Teams or Discord directly. You narrate punctuation out loud, saying "comma" and "period" where you want them. Full setup is in Microsoft's guide at support.microsoft.com.

2. 2

   On Mac: turn on Apple Dictation

   Open System Settings, then Keyboard, then Dictation, and switch it on. Trigger it with the dictation key (F5 on many MacBooks) or your own shortcut, click into the message box, and speak. Steps are in Apple's Dictation guide.

3. 3

   Fix the names before you hit Enter

   Read the message once. OS dictation nails plain sentences but stumbles on the exact tokens these apps live on, so scan for a mangled @mention, a wrong #channel, a broken emoji shortcode or code snippet, then correct it by hand before sending.

Where OS dictation falls apart in chat

Plain prose comes out fine. The trouble is that Teams and Discord messages are rarely plain prose. They are full of the exact things generic dictation mangles, and in a developer-heavy Discord server a single wrong symbol can change what a message means. Watch for these:

• @mentions and usernames, which come out as ordinary words instead of tagging the right person.
• #channel names, which lose the hashtag and the link.
• Emoji shortcodes like :shrug: or :tada:, which get spelled out as words.
• Inline code, backticks, and code blocks, which dictation has no way to represent.
• Technical jargon, library names, and product names, which get replaced with common-word guesses.
• Focus changes: if you click away or a notification steals focus, dictation can stop or type into the wrong place.

None of this makes the free route useless. For a quick "on my way, give me five minutes" it is perfect. It is the jargon-dense, handle-heavy message where you feel the seams.

If you want cleaner output: a fair roundup

If you dictate into chat all day, a dedicated system-level tool cleans up what Win+H and Apple Dictation leave rough, mainly punctuation you no longer narrate and a smarter guess at names. Importantly, none of these live inside Teams or Discord. They feed the same OS text field, just with better cleanup. So the honest framing is never "it integrates with Discord." It is "it types into the same box, cleaner."

Golem is one tool built this way. It works at the OS level, so it types into any focused field, including the Teams and Discord message box. It runs OpenAI's Whisper (large-v3-turbo) plus a cleanup pass that adds punctuation without you saying "comma," and it lets you teach it custom vocabulary, so the library names and handles you use constantly stop getting misheard. It is multilingual and can even output a different language than you speak. To be clear about its limits: Golem needs a login, and it needs an internet connection for the cleanest results because transcription runs server-side, so it is not an offline, on-device tool, and it runs on Windows and Mac only. It starts free with no credit card, and Pro is $3 a month or $30 a year.

Golem is not your only option, and it should not be the only one you weigh. Wispr Flow is a polished paid tool that turns rambling into tidy prose and runs on Mac and Windows, at a noticeably higher price of around $15 a month. Superwhisper is Whisper-based with a strong privacy and on-device focus, now on Mac, Windows, and iOS, with a free tier. Dragon by Nuance is the legacy powerhouse, expensive and enterprise-leaning. For developers, nerd-dictation is a free, open-source, scriptable option that leans Linux, and Talon Voice controls the whole OS by voice for accessibility, with a steep learning curve. Try the free OS route first, and only pay for cleanup if the daily friction justifies it.