← CantoVoice 粵聲

Cantonese Text to Speech, Done Properly

Most "multilingual" AI voice platforms — including ElevenLabs — don't natively support Cantonese. Reading Cantonese with a Mandarin model mangles the nine tones, chokes on colloquial characters like 嘅, 咗 and 啲, and stumbles over the English code-switching that defines real Hong Kong speech.

CantoVoice is built for Cantonese only. Our engine is trained on Hong Kong Cantonese, with a text-normalization pipeline that handles tones, dates and numbers, colloquial particles, and mixed English — so your voiceover sounds like a person from Hong Kong, not a machine guessing.

Who uses CantoVoice

• Cantonese YouTubers & content creators — narration for videos without recording yourself
• Overseas Cantonese communities (UK, Canada, Australia, US) — make content for the diaspora
• Businesses — product videos, training, IVR phone systems in natural Cantonese
• Educators & churches — lesson narration and sermon recordings

What you get

• Natural Hong Kong Cantonese voices with licensed voice talent
• MP3/WAV audio plus synced SRT subtitles for your video editor
• Per-word pronunciation override with Jyutping
• Commercial usage rights included in paid plans
• API access for developers (Business plan)

Pricing

Free 500-character trial. Paid plans from HK$68/month (~US$9) for 30,000 characters — versus HK$1,000+ for a single human voiceover session.

Join the waitlist — early birds get 40% off yearly →