Does Audiee.ai accurately transcribe Cantonese and mixed English–Chinese meetings?

Yes. Audiee.ai transcribes Cantonese, Mandarin and English in the same meeting — and in the same sentence — as they were spoken. Mixed mode is the default, so nobody picks one language before the call: a sentence that starts in Cantonese, borrows an English term and ends in Mandarin is written down the way it was said. A Cantonese transcript keeps the spoken words and particles — 佢哋, 唔係 and the rest — rather than tidying them into Mandarin; the summary is written in standard written Chinese; English terms stay in English. Thai, Vietnamese, Indonesian, Malay and Filipino with English mixed in are handled the same way. One thing Audiee.ai does not publish yet is an accuracy figure: the mixed-language benchmark is not out, so this page quotes no number and makes no promise that every word is right. The free plan gives you 3 hours of transcription a month to judge it on your own meetings.

Facts about Otter, Fathom, Notion and Fireflies checked September 2026 and linked to the pages they came from in Sources.

What a mixed Cantonese–English meeting looks like in Audiee.ai

No language to pick. Mixed mode is the default. Switch languages mid-sentence and the transcript follows; nobody has to decide before the call whether it is a Cantonese or an English meeting.
Cantonese is written as Cantonese. 佢哋 stays 佢哋 and 唔係 stays 唔係; colloquial words and particles are kept rather than translated into Mandarin equivalents. The guide on why Cantonese transcription drops words explains what usually goes wrong elsewhere.
English terms stay in English. A product name, a client's name or a technical term said in English inside a Cantonese sentence is written in English.
The summary is in standard written Chinese. The verbatim transcript keeps the spoken register; the summary, action items and what you are waiting on from others are written in the written register, each point with a timestamp that jumps to the transcript line.
Mandarin too. Mandarin and Cantonese in the same meeting are both transcribed; a sentence may move between them.

What other tools say about Cantonese and mixed languages

Tool	What its own pages say	In Audiee.ai
Otter	Six transcription languages; Chinese means spoken Mandarin, and Cantonese is not supported (checked September 2026).	Cantonese, Mandarin and English, mixed in one meeting.
Fathom	38 transcript languages, with no Chinese among them (checked September 2026).	Chinese and English together, including Cantonese.
Notion AI Meeting Notes	Business plan only; Cantonese support is not stated (checked September 2026).	Cantonese stated plainly, and on the free plan.
Fireflies	Its mixed-language mode is available on Business and above (checked September 2026).	Mixed mode is the default, on the free plan too.

None of this makes those tools bad; it is what their own pages listed when checked, and lists change. It is the reason to read the language list before relying on one of them for a Hong Kong, Singapore or Taiwan meeting that switches languages mid-sentence — see Audiee.ai vs Otter, Audiee.ai vs Fathom and Audiee.ai vs Fireflies for the full comparisons.

What Audiee.ai does not claim

No accuracy percentage. Until the mixed-language benchmark is published, Audiee.ai quotes no number and makes no promise about every word.
No automatic speaker names everywhere. Zoom names speakers automatically; on other platforms you rename a speaker once and it is remembered.
Not an offline tool. Audio is sent to a transcription service in the cloud and comes back as text. Audiee.ai's servers keep no copy, and the transcript and notes are Markdown files on your Mac — see what stays on your Mac and what is processed in the cloud.

How accurate is it? Is there a number?

Not yet. Audiee.ai has not published an accuracy figure because the mixed-language benchmark is not published, and a number quoted without it would be a guess. What you can check instead: every point in a summary carries a timestamp and jumps to the line it came from, and the transcript itself is never rewritten, so a doubtful line is one click from the audio. Mistakes happen — names and numbers most of all — so treat a transcript as something to check, not something to assume is perfect.

Does voice typing handle a mixed sentence the same way?

Yes. Tap Ctrl in any text box and speak in Cantonese, Mandarin and English as they come; the text arrives with filler words removed and punctuation added, and English terms stay in English inside a Chinese sentence. A word you correct once is recognized next time. Voice typing draws on the same monthly transcription hours as meetings, but it never counts as a meeting. Details on voice dictation on Mac.

Does it label who said what in a Cantonese meeting?

On Zoom, speakers are named automatically. On other platforms, rename a speaker once and Audiee.ai remembers the name from then on — so speaker names are not automatic everywhere, but the correction is a one-time job.

What does the free plan include?

3 hours of transcription a month, shared between meetings and voice typing, and 20 meetings. The knowledge base is not metered, and no card is needed to download. Pro is $19 a month, Ultra $49 — see pricing.

Does Audiee.ai handle Thai, Vietnamese, Indonesian, Malay or Filipino mixed with English?

Yes. A meeting in Thai, Vietnamese, Indonesian, Malay or Filipino with English mixed in is transcribed the same way as a Cantonese–English one — see Southeast Asian languages.

Sources

Download Audiee.ai — free, 3 transcription hours a month, no card.

Download Audiee.ai for macOS Free — 3 transcription hours / month · No card