Cantonese Speech to Text
LiveToka has a dedicated Cantonese speech recognition option, separate from Mandarin, so Cantonese speakers from Hong Kong, Macau, Guangdong and overseas communities can be transcribed and translated live.
New accounts get 10 free minutes. Sign in with Google — no credit card required.
Cantonese transcription: what to expect
- Cantonese is recognised as its own language — choose "Cantonese", not "Chinese (Mandarin)".
- Spoken Cantonese uses colloquial words that differ from written Chinese; the transcript reflects what is said.
- Translate into English, or into Mandarin (Simplified or Traditional) for readers who do not understand spoken Cantonese.
Popular uses
- Hong Kong business meetings with non-Cantonese colleagues
- Cantonese church services and community events overseas
- Interviews and oral-history recordings
- Following Cantonese livestreams and talks
How to start
- Click Start Cantonese transcription above and sign in with Google.
- Cantonese is pre-selected as the source language. Choose Transcribe Only, or Transcribe + Translate and a target language.
- Click Start Service and allow microphone access. Text appears as people speak.
Translating a call or livestream playing on your computer? See how to route meeting audio.
Other languages
- Mandarin & English speech to text
- Japanese speech to text
- Korean speech to text
- All supported languages
Frequently asked questions
Is Cantonese different from the Chinese (Mandarin) option?
Yes. They are separate recognition languages. Mandarin speakers should use "Chinese (Mandarin)" or the Mandarin & English bilingual mode.
Is there a free trial?
Yes. New accounts get 10 free minutes after signing in with Google — no credit card. Transcription uses credits at 1× and translation at 2×, so that is up to 10 minutes of transcription or 5 minutes of translation. Paid plans start at $9.9/month.
Do I need to install anything?
No. LiveToka runs in the browser. The latest Chrome, Edge or Safari is recommended.
Is my audio used to train AI?
No. Your voice recordings and transcripts are not used to train AI models.