Workers AI, Realtime - Real-time transcription in RealtimeKit now supports 10 languages with regional variants
Key Points
- 10 languages + regional variants
- Set ai_config.transcription.language at meeting creation
- Runs Nova-3 on Workers AI via AI Gateway — lower latency
Summary
RealtimeKit now provides real-time transcription powered by Deepgram Nova-3 running on Workers AI. Audio is routed through AI Gateway to Nova-3 on Cloudflare’s network for end-to-end processing, reducing latency versus external speech-to-text services. You can set the transcription language when creating a meeting; use multi for automatic multilingual detection.
Key Points
-
Supported languages: English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, Dutch.
-
Regional variants available (examples):
en-AU,en-GB,en-IN,en-NZ,es-419,fr-CA,de-CH,pt-BR,pt-PT. -
Set language in the meeting create payload (example):
{ "ai_config": { "transcription": { "language": "fr" } } }
-
Use
multito enable automatic multilingual detection. -
Ideal for voice agents and real-time translation workflows — no extra external STT services or routing logic required.
Actionable guidance
- When creating meetings, explicitly set
ai_config.transcription.languageto match caller language or usemultifor detection. - Expect lower latency when compared to routing audio off-network since transcription runs on Workers AI via AI Gateway.
- Verify regional variant codes for locale-specific transcription quality when required.