6 Best Cantonese Speech-to-Text Tools in 2026: Which Free Version Is Good Enough?

Many people dealing with Cantonese audio for the first time instinctively open their phone's built-in voice input or find a website that claims to offer free transcription, upload the file, and wait for the result.

But after one try, several common pitfalls become obvious: the free quota only covers the first few minutes, the model is actually Mandarin forced onto Cantonese, and Cantonese words like "m" "ge" "zo" all come out wrong. The output is just one long block of text with no punctuation, so you still have to listen again and retype everything yourself.

A more sensible approach is to first figure out what kind of recording you have—an online meeting, an in-person interview, or an already-recorded audio file—and then decide which type of tool to use. Below are 6 solutions that can actually handle Cantonese, along with how to use them and the easily overlooked details.

Why Do You Need Cantonese Speech-to-Text?

A 20-minute video, if you transcribe it by listening and typing, often takes an hour—and that's before fixing typos and syncing the timeline.

In daily work in Hong Kong and Guangdong, Cantonese meetings, client interviews, and YouTube video subtitles are common needs, but many transcription services only support English, and Cantonese is often categorized as "supported but not very usable."

This article lists the positioning, free and paid plans, and target users of 6 tools, and at the end explains a few things to note before actual use.

Tinrec (Miao Ting Lu Yin) — The Most Complete AI Recording Assistant

Tinrec is an AI meeting notes and collaboration tool for individuals and teams. It doesn't just turn sound into text; it organizes recordings into searchable, summarizable, queryable, and exportable data.

Record online meetings without a meeting bot. The desktop version directly captures your computer's system audio, handling platforms like Zoom, Google Meet, Microsoft Teams, and Webex. Clients or external attendees won't see an extra bot account, which is very practical for external meetings.

The real value comes after transcription. It automatically generates summaries, chapters, key points, and to-dos, and you can continue asking questions about the meeting content with AI Q&A, or generate reports, tables, and documents to fit right into your existing workflow.

Chinese and multilingual meetings. It covers Chinese, Traditional Chinese, and multilingual meeting scenarios, supports real-time translation and bilingual viewing, and lets you maintain company-specific terminology with team hotwords.

Team workspace. Meeting data belongs to the team, with roles and seats separating "who can manage the team" from "who can record, upload, and edit." When a member leaves, data can be handed over to others and won't disappear with their personal account.

For pricing, the personal plan includes a free version, weekly pass, Pro monthly, and Pro annual. Eligible team plans can try 7 days free with 1 free seat and 300 shared import minutes. Team monthly is USD 29.80 per paid seat per month, and annual is USD 199 per paid seat per year (about USD 16.58 per month), with each paid seat providing 2,000 shared team import minutes per month. Actual prices and currencies are subject to the official purchase page.

Best for office workers and teams that need Chinese meeting notes, don't want a bot joining the meeting, and still need to organize minutes and to-dos afterward.

Subanana — A Chinese Tool Centered on Cantonese Transcription

Upload an audio file or video, or paste a YouTube link directly, select Cantonese and the output format, and AI will produce a speaker-separated, punctuated transcript within minutes, handling mixed Chinese and English as well.

Output can be colloquial or written, Traditional or Simplified, and exported as SRT, VTT, TXT, DOCX, XLSX, or Markdown. The first 15 minutes of each file can be previewed for free; subsequent plans are subject to the official website.

Best for those focused on Cantonese video subtitles and interview transcripts, who don't yet need meeting management and team collaboration.

CantoSub AI — A Transcription Service Specializing in Cantonese Variants

Upload audio or video and it converts Cantonese speech into Traditional Chinese text, supporting MP4, MOV, MP3, WAV, and other formats.

Its specialty is supporting multiple Cantonese variants such as Hong Kong Cantonese, Guangzhou dialect, Dongguan dialect, and Shunde dialect, and it offers personalized dictionary settings. It can also automatically convert colloquial Cantonese into written Chinese and export .txt, .docx, .pdf. Pricing is subject to the official website.

Best for Hong Kong and Guangdong users with heavier accents or teams where everyone speaks a bit differently.

Google AI Studio (Gemini) — A Zero-Cost Temporary Solution

Google AI Studio with Gemini can directly transcribe audio to text, supports desktop and mobile, requires no additional API setup, performs well on Cantonese, and can handle long content.

After transcription, you can ask it to organize, summarize, or analyze the transcript in the same conversation, which is convenient for those who only occasionally handle recordings. It's free to use, but lacks a project-based meeting management interface, and files and records aren't automatically archived.

Best for those who only use it a few times a month and don't want to sign up for another new service.

Notta — Multilingual Transcription and Meeting Notes

Notta covers multi-platform transcription, meeting notes, file import, and translation, with multilingual support as its strength.

The free version offers 120 minutes per month, with a maximum of 3 minutes per item; Pro annual is about USD 8.17 per month with 1,800 minutes per month; Business annual is about USD 16.67 per seat per month, offering unlimited transcription, role permissions, and usage reports. Pricing is subject to Notta's official pricing page.

Best for those already used to English-interface tools and needing mixed-language transcription.

Speechnotes Hong Kong Cantonese — A Pure Online Free Notepad

Speechnotes' Hong Kong Cantonese version is an online speech-to-text notepad; open the webpage and you can speak while it transcribes.

It has built-in punctuation and Cantonese character correction, converting half-width punctuation to full-width and fixing Cantonese words spelled with "o" into single characters, saving a lot of proofreading time.

Best for short, real-time dictated notes or drafts; long meeting recordings are not its strong suit.

With So Many Options, How Should You Choose?

• Want to get summaries, chapters, key points, and to-dos right after transcription → Tinrec (transcription is only half its function)
• Don't want an extra bot joining online meetings → Tinrec (desktop version captures system audio)
• Need Chinese, Traditional Chinese, and multilingual mixed meeting content → Tinrec
• Need to generate reports, tables, or documents after the meeting → Tinrec (Agent post-processing)
• Team needs to share meeting data, manage seats and usage → Tinrec Team
• Need to continue between phone, tablet, and computer → Tinrec (available on multiple platforms)
• Just occasionally transcribe a Cantonese recording or YouTube video → Subanana, CantoSub AI
• Use less than a few times a month and don't want to spend money → Google AI Studio (Gemini)

5 Things to Know Before Using Cantonese Speech-to-Text

1. Recording quality determines half the result. Too far away, heavy echo, multiple people talking at once—any tool will make mistakes. Placing your laptop or phone in the center of the table during a meeting saves much more effort than fixing typos afterward.

2. Decide colloquial or written first. The same passage transcribed as colloquial (keeping "m" "ge" "zo") versus written serves completely different purposes. Subanana and CantoSub AI both offer output options; before starting, decide whether this text is for subtitles or meeting minutes.

3. Don't treat free quotas as unlimited. Most services have limits on single file length or monthly minutes. For example, Tinrec Team provides 2,000 shared import minutes per paid seat per month. Before choosing, estimate how long you'll record in a month.

4. AI transcription is not proofreading. Names, product names, and technical terms are most error-prone. Tools like Tinrec can maintain common terms via hotwords; for important content, it's still recommended to verify against transcript timestamps and audio playback.

5. Confirm consent and regulations before recording. Before recording a meeting or interview, remember to inform participants and obtain consent, and comply with local laws and regulations.

FAQ

Q1: Is a free tool enough for Cantonese speech-to-text?
It depends on frequency. For occasional recordings, a free option like Google AI Studio is enough; if you have meetings every week, free quotas run out quickly, and you end up wasting time switching tools.

Q2: Can I transcribe directly on my phone?
Yes. Most services have apps or mobile web versions. But if you're recording an online meeting, using a computer desktop version to capture system audio is much clearer than pointing your phone at the speaker.

Q3: Can the transcribed Cantonese colloquial text be used directly?
It depends on the purpose. For subtitles, keeping colloquial is usually more natural; for meeting minutes or reports, it's better to choose written output or have the tool organize it again after transcription.

Q4: Can mixed Chinese-English meetings be handled?
Yes. Subanana states it reliably handles mixed Chinese-English, and Tinrec also covers Chinese and multilingual mixed meeting scenarios.

Q5: How should I record online meetings?
If you don't want a bot account appearing in the meeting room, you can choose a desktop version that captures system audio, such as Tinrec. Users simply start recording themselves without adding an extra account to the meeting.

Q6: For team use, should I choose personal or team version?
If meeting data needs to be shared, with member and seat management, and usage visibility, choose the team version. Tinrec Team provides an independent team space, meeting data belongs to the team, and includes roles, seats, usage analytics, and audit logs.

Conclusion

The barrier to Cantonese speech-to-text isn't high; the hard part is turning that long block of text into useful data after transcription.

If you only occasionally need a transcript, Subanana, CantoSub AI, or Google AI Studio can all solve it. But if you have meetings every week and need to organize minutes, track to-dos, and keep data for the team afterward, a tool like Tinrec that connects transcription, summarization, Q&A, export, and team workspace will save much more time.

Try it with a Cantonese recording you already have, compare the transcription quality and organization cost, and then decide whether to use it long-term.

References

• Free + Cantonese OK: How to use Google AI Studio to turn your recording into ...
• [Free Tool] Cantonese Speech to Text! Just upload a file or YouTuber link. Up to 96% accuracy - YouTuber AI | Spend time on creativity, leave the AI to do the hustle work.
• Free Cantonese Speech to Text App 2026 | Subanana
• Speechnotes | Hong Kong Cantonese | Speech to Text Online Notepad
• AI Cantonese Speech to Text Professional Cantonese Recognition