How to Use Cantonese Voice Typing in Windows 11 (2026): Setup Guide + Meeting Transcription Tools

Many people think Windows 11 doesn't have Cantonese voice typing—but you actually have to turn it on yourself

Lots of people try voice typing, say a couple of sentences, see it only recognizes written Chinese, and conclude that "Windows doesn't support Cantonese." In reality, the problem usually isn't that the system doesn't support it—it's that the language and speech packs aren't installed, or the voice typing language hasn't been switched over.

Another common inefficient approach is treating voice typing as a meeting notes tool. During a meeting, you talk into the microphone and watch text scroll by line after line. After the meeting, you realize those words weren't saved to a file, there's no summary, and you can't look up who mentioned the budget last week.

If you're stuck on either of these issues, this article will first help you set up Cantonese voice typing in Windows 11, then discuss what it can and can't do, and what tool you should use for meeting notes.

Before you start setting up, understand these 4 key points

1. Voice typing uses online recognition and requires an internet connection. Windows voice typing uses online speech recognition powered by Azure Speech Services. If the internet goes down, this feature is basically unusable.

2. Cursor position matters more than anything. It only types into the text box where the cursor currently is. If the cursor isn't in an input field, no matter how much you speak, no text will appear.

3. Language and speech packs must be installed separately. To recognize spoken Cantonese, you need to add the corresponding language in Language & Region settings and download the speech-related packs. If you only keep the default Chinese, it will recognize written Chinese.

4. Environment and microphone determine recognition quality. Voice typing has echo cancellation, which automatically ignores sound coming from speakers (such as Narrator or app playback) and only listens to your voice. Conversely, in a noisy environment or with the microphone too far away, recognition quality drops noticeably.

There's also an often-overlooked setting: the wait time before acting. It determines how long voice typing waits before executing voice commands, and both slow and fast speakers can adjust it to their preference.

How to enable Cantonese voice typing in Windows 11: 5 steps

Step 1: Connect your microphone first. Whether it's a laptop's built-in mic, a headset mic, or a USB microphone, go to system settings first to confirm it's detected.

Step 2: Add the language you want to recognize. Go to Settings → Time & Language → Language & Region, add the Cantonese-related language item (the option name varies by Windows version, e.g., Chinese (Hong Kong)), and after adding, remember to download the speech pack for that language.

Step 3: Switch input language. Press Windows logo key + Spacebar to switch between installed languages, and confirm the currently selected one is the one you want to use for Cantonese recognition.

Step 4: Place the cursor in a text box and bring up voice typing. In Windows 11, the common method is pressing Windows logo key + H. A microphone icon for voice typing will appear at the top of the screen; click it to start speaking. Shortcuts may vary slightly by version, and you can also confirm from voice typing settings.

Step 5: Start speaking and enable auto punctuation as needed. Besides simple input, you can also use simple voice commands, such as deleting words or selecting sentences. To save time on manually adding punctuation, you can enable automatic punctuation in settings.

Also, if you're using a Copilot+ PC, keep an eye on the new "Fluid Dictation" feature. It automatically corrects grammar, punctuation, and filler words as you speak, processed by an on-device small language model for speed and privacy. When you enable voice typing, the system automatically downloads the model in the background; before the download completes, the Fluid Dictation option in settings will be grayed out. To check progress, go to Settings → Windows Update to see the download status of the AI Phi Silica model.

It's useful, but don't treat built-in voice typing as a meeting notes tool

Using Windows 11's Cantonese voice typing as a "typing replacement" works quite well. But its role is an input method, not a recording tool.

Here are things it inherently cannot do:

• It doesn't save audio recordings—once you finish speaking, it's gone.
• It doesn't automatically create a transcript file; text is scattered in whatever document you were typing in.
• It doesn't generate summaries, chapters, or action items.
• It can't upload existing meeting recordings or videos and convert them to text.
• It can't search or share across devices or team members.
• The cursor must be in a text box for text to go in.

In other words, it's great for real-time typing; but for questions like "What exactly did the client promise in that meeting three months ago?" it can't help. That's where meeting notes tools come in.

Tinrec: Turn spoken words into searchable, collaborative meeting data

Tinrec is an AI meeting notes and collaboration tool for individuals and teams, available on iOS, Android, desktop, and web. It doesn't just convert audio to text—it organizes meeting recordings into searchable, summarizable, queryable, exportable, and further processable data.

In practice, it works like this: during a meeting, you start live recording, and it produces a transcript as it records. When the meeting ends, the AI summary and chapters are already organized, and it picks out action items from the discussion—who's responsible for what and when it's due, at a glance.

For online meetings, you don't need to invite an extra bot. Tinrec's desktop version directly captures your computer's system audio, handling meetings on Zoom, Google Meet, Microsoft Teams, Webex, and similar platforms without an extra bot joining the participant list. For those who "don't want clients to see a stranger's account in the meeting room," this is especially appreciated.

What's even better is post-meeting. After recording, you can directly ask questions about the meeting content, such as "Who mentioned the budget in the last meeting?" It gives you answers based on semantic understanding, rather than throwing a string of keywords for you to dig through. The same data can also be exported to tools like Notion, Google Docs, OneNote, Dropbox, or further generate reports, tables, and meeting minutes.

Worth mentioning is the team version. It provides an independent team space where meeting data belongs to the team, and you can manage member roles, seats, sharing rights, usage analytics, and audit logs. When a member leaves or is removed, the data isn't taken away but handed over to other valid members. For teams where "meeting notes are scattered across everyone's computers," this is the most practical difference.

On pricing, the personal version starts with a free tier offering basic quota to experience recording, import, summary, and transcription; for short-term intensive meetings or interviews, consider a weekly pass; for long-term stable use, Pro monthly or annual plans are the main choice. For the team version, first-time eligible teams can try 7 days, 1 free seat, and 300 minutes of team shared import quota; team monthly is USD 29.80 per paid seat, annual is USD 199 per paid seat (about USD 16.58 per month), with each paid seat providing 2,000 minutes of team shared import quota per month. Live recording under an active team plan currently doesn't deduct minutes. Prices and benefits are subject to the official purchase page.

One reminder: Tinrec focuses on Chinese and multilingual meeting scenarios, and actual transcription quality is affected by recording environment, accents, multiple people speaking simultaneously, and technical terminology. It's not recommended to understand it as a "fixed 98% accuracy rate." If you mainly record Cantonese meetings, it's advisable to first use the free quota to test with a segment of your own meeting recording before deciding.

Who is it for? If you just need to type a few hundred words of messages daily, Windows built-in voice typing is plenty. But if you need to record meetings, later use AI to ask follow-up questions, or even let the whole team share the same meeting data, Tinrec is the type of tool that better matches those needs.

Besides built-in voice typing and Tinrec, what other options are there?

VoiceIn: A Chrome extension that uses Google's speech recognition API under the hood, supports Cantonese, and can handle mixed Chinese-English sentences, working in text boxes on any website. Suitable for people who "only write in the browser." However, it doesn't save recordings, generate meeting summaries or action items, and has no team space—these are exactly Tinrec's strengths.

Built-in transcription in Zoom, Google Meet, Teams: If you only meet on a single platform, turning on the platform's built-in feature is easiest. The downside is data is scattered across platforms, and switching meeting software means learning it all over again; Tinrec collects across platforms, centralizing transcripts, summaries, and action items in one place.

Otter.ai: A mature choice for English meetings, with a free plan of 300 minutes per month—quite generous. But it doesn't support Cantonese, and Chinese meetings aren't its strong suit. If almost all your recordings are in English, Otter.ai is a more cost-effective option than Tinrec; when you need Chinese and bot-free recording, Tinrec is more suitable.

Notta / TurboScribe: If you have a bunch of existing audio or video files that need batch conversion to subtitles or text, these tools have advantages in long-file processing. Tinrec's strength isn't in one-time bulk files, but in live meetings, post-meeting organization, AI Q&A, and team collaboration.

Pitfall guide: 4 most common mistakes in setup and tool selection

Pitfall 1: Assuming you can speak Cantonese right after the system is installed. If the language and speech packs aren't separately installed and switched over, it will recognize written Chinese. First confirm the language list and speech packs are in place, then start testing.

Pitfall 2: Starting to speak without placing the cursor in an input box. This is why many people speak for five minutes in vain. First click into the field you want to type in, see the cursor blinking, then start voice typing.

Pitfall 3: Testing in a noisy environment and then blaming the tool for inaccuracy. Whether it's Windows built-in or any transcription service, environmental noise, microphone distance, and overlapping speakers directly affect results. Test with your real scenarios, not just official demos.

Pitfall 4: Confusing the two types of tools. Voice typing solves "don't want to type," meeting notes tools solve "need to use it after the meeting." Using the former for the latter's job only leaves you with a screen full of text but no conclusions.

One final reminder unrelated to tools but important: recording and transcription involve participants' rights, and regulations vary by region. Before recording, please confirm whether you need to obtain the other party's consent.

Conclusion: Which one should you choose?

In one sentence: use Windows 11 built-in voice typing for daily typing, and leave meeting notes to Tinrec.

• Want to replace the keyboard with your voice for typing Chinese in LINE, Word, or any browser input box → Windows 11 built-in Cantonese voice typing, free, no extra installation.
• Need to record online meetings without a bot joining the meeting room → Tinrec desktop's bot-free recording.
• After recording, want to directly ask AI "What was the conclusion of this meeting?" "Who's responsible for which item?" → Tinrec's AI Q&A and action item extraction.
• Meeting data needs to be shared with the whole team, and you need to manage members, seats, and usage → Tinrec team version.
• Only typing text in the browser, zero budget → Extensions like VoiceIn are worth trying.
• Almost only recording English meetings → Otter.ai's free quota is more cost-effective.

Suggested order: First set up Windows built-in Cantonese voice typing and get a feel for its convenience and shortcomings; then use Tinrec's free quota to record a segment of your own meeting and see if the transcript and summary are sufficient. If it works for you, consider upgrading—no need to pay from the start.

References

• So Windows 11 finally has built-in Cantonese voice typing | LIHKG
• Use voice typing to talk instead of type on your PC | Microsoft Support
• A detailed guide to Windows voice typing for first-time users
• VoiceIn Cantonese speech recognition: type three times faster on desktop without typing
• Windows 11 Series EP.04 Built-in Cantonese voice typing chat | Computer Research Institute...