The Soniox ASR integration allows AVR to leverage Soniox’s high-accuracy, low-latency real-time speech recognition via WebSocket streaming.
Soniox is well known for delivering state-of-the-art transcription quality, outperforming most cloud providers across noisy environments, accents, and spontaneous speech.
Official website: https://soniox.com/
Real-time ASR: https://soniox.com/docs/stt-realtime
Soniox delivers several strong advantages, making it an excellent ASR choice for real-time voicebots:
Clone the official Soniox ASR integration:
git clone https://github.com/agentvoiceresponse/avr-asr-soniox.git
| Variable | Description | Default | Required |
|---|---|---|---|
SONIOX_API_KEY |
Your Soniox API key | — | Yes |
SONIOX_WEBSOCKET_URL |
Soniox real-time WebSocket endpoint | wss://stt-rt.soniox.com/transcribe-websocket |
Optional |
SONIOX_SPEECH_RECOGNITION_MODEL |
ASR model to use | stt-rt-v3 |
Optional |
SONIOX_SPEECH_RECOGNITION_LANGUAGE |
Language code or comma-separated list | en |
Optional |
PORT |
Local service port | 6018 |
Optional |
.envSONIOX_API_KEY=your_soniox_api_key
SONIOX_WEBSOCKET_URL=wss://stt-rt.soniox.com/transcribe-websocket
SONIOX_SPEECH_RECOGNITION_MODEL=stt-rt-v3
SONIOX_SPEECH_RECOGNITION_LANGUAGE=en
PORT=6018
avr-asr-soniox:
image: agentvoiceresponse/avr-asr-soniox
platform: linux/x86_64
container_name: avr-asr-soniox
restart: always
environment:
- PORT=6018
- SONIOX_API_KEY=${SONIOX_API_KEY}
- SONIOX_WEBSOCKET_URL=${SONIOX_WEBSOCKET_URL:-wss://stt-rt.soniox.com/transcribe-websocket}
- SONIOX_SPEECH_RECOGNITION_MODEL=${SONIOX_SPEECH_RECOGNITION_MODEL:-stt-rt-v3}
- SONIOX_SPEECH_RECOGNITION_LANGUAGE=${SONIOX_SPEECH_RECOGNITION_LANGUAGE:-en}
networks:
- avr
Set ASR_URL in your AVR Core configuration:
ASR_URL=http://avr-asr-soniox:6018/speech-to-text-stream
Once set, AVR Core will:
stt-rt-v3 model (default)