6 Free Cantonese Speech-to-Text Tools Tested: Which Is Best for Cantonese Transcripts in 2026?

Most people searching for "free Cantonese speech-to-text" make their first mistake by assuming free means no payment required.

The second mistake is more subtle: treating text-to-speech tools as speech-to-text.

I recently organized a batch of Cantonese interview recordings and realized these two things are often mixed up in search results.

Most free options aren't fully featured—they're "trial quotas."

So the real question isn't "which one is free," but: what am I transcribing? Where does my workflow end?

If you just want to turn a 5-minute recording into text, free quotas are often enough.

But if that transcript later needs to become a summary, action items, or subtitles, what you need is a workflow, not free.

This article covers 6 options. I'll start with the one I use most, then cover 5 alternatives.

Why do you need Cantonese speech-to-text?

Manually subtitling a 20-minute Cantonese video often takes over an hour.

The real challenge isn't the speech—it's the text: should "唔係" (not), "嘅" (possessive), "咗" (completed action) be converted to written Chinese or kept as-is?

That depends on what you'll do with the transcript afterward.

So before choosing a tool, clarify your purpose: is it for yourself, for your boss, or for direct subtitling?

Option 1 | Tinrec — The most complete AI meeting notes and post-meeting workflow

In one sentence: turn recordings into data you can keep working with, not just a transcript.

Highlight 1: The work starts after recording

Transcription is just the beginning. Tinrec automatically generates summaries, chapters, key points, and action items, turning long meetings into readable segments—you can grasp the whole meeting in five minutes.

Highlight 2: Ask follow-up questions about meeting content

This is what sets it apart from other transcription tools. Transcripts are often just "records," but Tinrec lets you ask questions about a single meeting or across meetings, turning dead data into data you can mine.

Highlight 3: No bot needs to join the meeting

The desktop version captures system audio directly—no need to invite an extra participant to Zoom, Google Meet, Microsoft Teams, or Webex. Recording is initiated by you, not automatically joining all calendar meetings in the background.

Highlight 4: Data stays

Historical meetings are centralized in one place, searchable, queryable, and exportable to Notion, Google Docs, OneNote, or Dropbox for continued use. They don't disappear after the meeting.

(Screenshot: Tinrec workspace showing transcript and summary side by side, with an arrow pointing to the AI Q&A input box)

Free and paid plans

The free version offers basic quota to experience recording, import, summaries, and transcription; there are also weekly cards, Pro monthly, and Pro annual plans to choose from based on usage frequency. The team version offers a 7-day trial (1 free seat, 300 minutes of shared team import quota), USD 29.80 per paid seat per month, or USD 199 per paid seat per year; paid seats include 2,000 minutes of shared team import quota per month.

Let's be honest about language support

Tinrec covers Chinese, Traditional Chinese, and multilingual transcription, and supports mixed Chinese-English content organization and real-time translation. As for actual performance with Cantonese accents and mixed Chinese-English, it will be affected by recording quality, ambient noise, overlapping speakers, and technical terminology—we recommend running a segment of your own real recording with the free quota first before deciding.

Who it's for

People who need post-meeting outputs, long-term recording retention, or team sharing of meeting data.

Option 2 | Subanana — Free quota transcription specialized for Cantonese

Clear positioning: a transcription service centered on Cantonese, with the first 15 minutes of each file free to preview, no credit card required.

It has targeted handling for mixed Chinese-English, keeping English words in sentences as-is; output can be colloquial or written, with four combinations of Traditional and Simplified Chinese, plus speaker identification, custom vocabulary lists, and YouTube link import.

Limitations to note: free is "preview transcription results"; exporting files is a paid feature.

Best for those with pure Cantonese content who want to first confirm "whether it transcribes correctly."

Option 3 | cSubtitle — Transcript and subtitles in one go

Positioning: convert audio files to transcripts and also subtitle videos.

It works for meeting notes, interviews, courses, lectures—any scenario requiring "textification," with a free trial.

Best for those who need both transcript and subtitle files.

Option 4 | YouTuber AI — Paste a link and transcribe

Positioning: upload a file or paste a YouTube link to transcribe.

Its page claims up to 96% accuracy—that's their own claim; actual results will vary with recording quality, accent, and background noise. I treat it as a reference, not a guarantee.

Best for those whose source material is already on YouTube and don't want to download files first.

Option 5 | Notta — Multilingual transcription and translation

Positioning: multi-platform transcription, meeting notes, file import, translation, and team plans—a broad-coverage player in this space.

Free version: 120 minutes per month, but max 3 minutes per session—this limit significantly affects long meetings. Pro annual billing is about USD 8.17/month with 1,800 minutes per month; Business annual billing is about USD 16.67 per seat per month.

Best for those who need multilingual translation and can accept the free version's 3-minute per-session limit.

Option 6 | TurboScribe — Long files and batch uploads

Positioning: a cost-effective file transcription route, focusing on long documents, batch uploads, subtitles, and translation.

Free version: 3 files per day, max 30 minutes per file; Unlimited annual is USD 120 (about USD 10/month), paid version supports up to 10 hours or 5 GB per file, and up to 50 files at once.

Best for those with lots of long recordings who don't need real-time meeting notes.

With so many options, how should you choose?

• Want summaries, chapters, action items after transcription, and even follow-up questions about meeting content → Tinrec
• Need to keep original audio for later verification → Tinrec (saves recordings; some bot-free note tools on the market don't keep meeting audio)
• Need to take notes during online meetings without adding a bot → Tinrec desktop version
• Content is mainly colloquial Cantonese and you want to verify accuracy first → Subanana's first 15 minutes free preview
• Need subtitle files, not just transcripts → cSubtitle
• Source material is already on YouTube and you don't want to download → YouTuber AI
• Need to process lots of long recordings at once → TurboScribe
• Zero budget, only occasional transcription → Use each tool's free quota first, don't rush to pay

(Screenshot: Self-made comparison table of free quotas across options, with arrows marking "per-session limit" and "export fees" columns)

4 things to know before using Cantonese speech-to-text

1. Free usually means "quota," not "features"

Subanana free is preview, Notta free is 3 minutes per session, TurboScribe free is 3 files per day. Check the limits first—that's what makes your comparison meaningful.

2. Text-to-speech is not speech-to-text

Online tools like Narakeet and Piliapp are TTS: you input text, it reads it aloud. The direction is completely opposite—don't pick the wrong tool in search results.

3. Decide on colloquial vs. written style first

For subtitles, keep colloquial; for meeting notes, convert to written. If you don't think about this first, you'll have to redo everything later.

4. Recording quality determines everything

Multiple people talking over each other, noisy environments—even the best model will struggle. For important content, we recommend listening back to the original recording along the timeline for verification; before recording, don't forget to inform participants and comply with local regulations.

Conclusion: Build your workflow first, then choose tools

Tools are never the problem—the workflow is.

My advice: take a segment of your own real recording and run it through each option's free quota. You'll quickly discover which one's output you can use directly.

Quick start checklist

1. Pick a 10-minute real recording with mixed Chinese-English as a test sample
2. Decide first whether you want colloquial or written style for this transcript
3. Run it with free quotas—don't pay first
4. Check only three things: proper nouns, speaker segmentation, punctuation
5. Save confirmed terminology as hotwords or a vocabulary list for next time
6. If you need long-term retention and post-meeting outputs, consider a full-workflow solution like Tinrec

Run a round first, then decide whether to spend money. Test step by step, and you'll gradually find your own approach.

References

• Free Cantonese Speech-to-Text App 2026 | Subanana
• Cantonese Text-to-Speech
• Text-to-Speech Cantonese
• [Free Tool] Cantonese Speech-to-Text! Just upload a file or YouTube link. Up to 96% accuracy - YouTuber AI | Spend time on creativity, leave the AI to do the hustle work.
• AI Speech-to-Text Fast Audio File Online Transcription, Transcripts and Video Subtitles - Free Trial - cSubtitle