5 Cantonese Speech-to-Text Apps Tested for 2026: Which Handles Mixed Chinese-English Best?

Why Do You Need a Cantonese Speech-to-Text App?

Re-listening to a one-hour meeting to type up a transcript often takes two or three times as long; the same goes for interviews, classes, and client calls. Cantonese transcription is particularly difficult: there are many colloquial characters and frequent Chinese-English code-switching. Sentences like "this quarter's target" can easily come out garbled if you force a Mandarin model on them.

That's why most people search for "is a Cantonese speech-to-text app useful"—it's not that there are no tools, but they worry the output won't be usable and they'll have to edit it from scratch. Below are 5 practical options, covering free and paid, cloud and offline, software and hardware, followed by selection advice.

Tinrec (Miao Ting Recording) — Turn Meetings into Searchable, Collaborative Team Data

Tinrec is an AI meeting notes and collaboration tool for individuals and teams. It's positioned not just as "speech to text" but as a way to organize recordings into searchable, summarizable, queryable, and exportable data.

Record online meetings without a meeting bot. The desktop version directly captures computer system audio, handling Zoom, Google Meet, Microsoft Teams, Webex, and other online meetings without needing an extra bot to join as a participant. Recording is started manually by the user, making it suitable for situations where you don't want a "Tinrec Bot" added to the meeting list.

The real value comes after recording. The system automatically generates summaries, chapters, key points, and action items. You can ask AI follow-up questions about meeting details, such as "What was the supplier's quote?" without re-listening to the entire recording. You can also generate reports, tables, and meeting minutes, or export to Notion, Google Docs, OneNote, Dropbox, and other tools for further processing.

Chinese and multilingual meetings. Supports Chinese meeting transcription, mixed Chinese-English content organization, and multilingual transcription. During recording, you can view in original text, translation, or bilingual mode, so cross-language meetings don't require post-meeting translation. The team version also offers hotword management to capture company-specific terminology and product names.

Team spaces and governance. Meeting data belongs to the team; new members can view historical meetings. Administrators can manage roles and seats, view usage trends and details, query and export audit logs. When a member leaves, resources can be transferred to other active members and won't disappear with them.

Plans and pricing. The personal version includes a free tier (basic quota for recording, import, summaries, and transcription), a weekly pass, a Pro monthly plan, and a Pro annual plan—choose based on usage frequency. The team version offers a 7-day trial (1 free seat, 300 minutes of shared team import quota). Monthly billing is USD 29.80 per paid seat; annual billing is USD 199 per paid seat (about USD 16.58/month). Paid seats include 2,000 minutes of shared team import quota per month. Under an active team plan and seats, real-time recording currently does not deduct minutes. Actual prices and benefits are subject to the official purchase page.

Who it's for: Office workers who have daily meetings and need to submit meeting minutes afterward; department heads and project teams who need to turn cross-department meetings into team assets.

Subanana — Focused on Cantonese Models and Spoken-to-Written Conversion

Positioned as a purely Cantonese-oriented transcription tool. Upload audio or video files, or paste a YouTube link, and AI converts Cantonese dialogue into a transcript with speaker separation and punctuation. English words within sentences are kept as-is, not converted to phonetically similar Chinese. Output can be colloquial or written; when written is selected, "佢哋而家搞掂咗" becomes "他們現在處理好了". Supports common formats like MP3, M4A, WAV, MP4, MOV. The first 15 minutes of each file can be previewed for free, ideal for testing your own recording before deciding to pay. Suitable for those who need Cantonese verbatim transcripts and want written output directly.

Notely Voice — 100% Offline Open-Source Option

An open-source application focused on 100% offline processing, with built-in Whisper model, supporting Cantonese and automatically converting spoken to written language. The biggest selling point is that recordings don't need to be uploaded to the cloud, appealing to those who value privacy and don't want to hand internal meeting recordings to third-party servers. Suitable for sensitive content or those who often work in unstable network environments.

Plaud Note / NotePin — AI Recording Hardware Route

If you need "portable recording" rather than software, Plaud is a hardware option to consider. It supports 112 languages including Chinese (Mandarin/Cantonese). After purchasing the device, you get 300 minutes of transcription service per month, with additional paid membership plans. The advantage is record-on-the-go, suitable for client visits, ad-hoc interviews, phone recordings, etc.; the downside is carrying an extra device, and hardware and membership are billed separately.

iPhone Built-in Recording + Google Docs — Zero-Cost Starter Combo

iPhone's built-in speech recognition takes a "faithful reproduction" approach, preserving original colloquialisms like "喺" and "唔使". General vocabulary recognition is quite stable, but it doesn't automatically convert to written language, nor does it produce summaries or action items. Google Docs and Google Keep voice input are only suitable for short real-time dictation; they don't support uploading audio files and can't distinguish speakers. Suitable for those with zero budget who only occasionally need to turn a passage into text.

With So Many Options, How Should You Choose?

• Want data you can use directly after transcription (summary + chapters + action items + AI Q&A) → Tinrec, one of the few solutions that covers the entire post-meeting workflow.
• Need mixed Chinese-English meeting minutes and want to build a team knowledge base → Tinrec, supports mixed Chinese-English content organization and centralized team space management.
• Don't want an extra bot joining online meetings → Tinrec desktop version, directly captures system audio.
• Switch between phone, computer, and web → Tinrec, multi-device use with data sync.
• Just want a pure Cantonese transcript and prefer written output → Subanana, try the free 15-minute preview first.
• Recording content is sensitive and must not go through the cloud → Notely Voice's offline processing.
• Often on the go and need a physical device for quick recording → Plaud Note / NotePin.
• Don't want to spend any money and only use occasionally → iPhone built-in recording or Google Docs voice input.

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

1. Test with your own recording first. The same tool can perform very differently in a quiet meeting room versus a noisy cha chaan teng. Subanana offers a free 15-minute preview per file, and Tinrec has a free tier quota. Testing with a real recording is more accurate than any review.

2. Mixed Chinese-English is the biggest challenge. Check whether English words have been forcibly translated into Chinese and whether proper nouns are distorted. This is especially critical in industries with frequent bilingual meetings and is the most direct indicator of a tool's quality.

3. Decide whether you want a colloquial or written transcript. For archiving, publishing, or client-facing documents, choose a tool that outputs written language; for verbatim reproduction of on-site dialogue for research or legal purposes, keeping colloquial language is more faithful.

4. Audio quality sets the ceiling. Environmental noise, multiple people talking over each other, and being too far from the microphone will noticeably reduce accuracy. Keeping the phone or recording device as close to the speaker as possible is the lowest-cost improvement. Also, when recording involves others, be mindful of local regulations and obtain consent from participants when necessary.

Conclusion

The difficulty of Cantonese speech-to-text lies in mixed Chinese-English and colloquial recognition, not just speech-to-text conversion. If you only want a transcript, Subanana and Notely Voice each have their strengths; but if you want "something usable right after the meeting"—summaries, action items, queryable meeting data, or even a team-wide meeting database—Tinrec is the most complete workflow in this list. Try it with your own recording using the free quota before deciding to pay; that's more practical than subscribing blindly.

References

• Free Cantonese Speech-to-Text App 2026 | Subanana
• Tutorial: Free AI Offline Speech-to-Text App Cantonese Spoken-to-Written Conversion, Comparison with iPhone Built-in Transcription
• Top 5 Speech-to-Text Tools in Hong Kong: Accurate Cantonese Recognition, Double Meeting and Interview Efficiency!
• SoundType AI Ultra-Accurate Transcripts, Speech-to-Text, AI Dictation - Google Play App
• 2026 Latest 4 Cantonese Speech-to-Text Tools: Solving Mixed Chinese-English and Meeting Notes Pain Points - Miao Ting Recording Blog