¶ Using Vosk ASR with AVR
Vosk is an open-source speech recognition toolkit that runs locally without requiring internet connectivity. It supports more than 20 languages and dialects and is a great choice if you need offline, privacy-friendly transcription.
¶ Supported Languages
Vosk provides pre-trained models in over 20 languages, including:
English, Indian English, German, French, Spanish, Portuguese, Chinese, Russian, Turkish, Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino, Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech, Polish, Uzbek, Korean, Breton, Gujarati, Tajik, Telugu.
Full list and downloads available here: Vosk Models
¶ Docker Setup
To use Vosk with AVR, add the following service to your docker-compose.yml:
avr-asr-vosk:
image: agentvoiceresponse/avr-asr-vosk
platform: linux/x86_64
container_name: avr-asr-vosk
restart: always
environment:
- PORT=6010
- MODEL_PATH=model
volumes:
- ./model:/usr/src/app/model
networks:
- avr
¶ Example: AVR Core + Vosk
avr-core:
image: agentvoiceresponse/avr-core
container_name: avr-core
restart: always
environment:
- ASR_URL=http://avr-asr-vosk:6010/speech-to-text-stream
...
networks:
- avr
¶ Environment Variables
Variable
Description
Example Value
PORT
Port on which the Vosk ASR service runs
6010
MODEL_PATH
Path to the mounted Vosk model (container)
model
¶ Performance Notes
- Small models → Faster, require fewer resources, but less accurate
- Large models → Higher accuracy, but demand more CPU/RAM/GPU
For most AVR use cases, small/medium models provide a good balance between latency and accuracy
¶ Troubleshooting
- Error: No model found → Ensure the model directory is correctly mounted and not empty
- High CPU usage → Switch to a smaller model or allocate more resources
- Slow recognition → Verify Docker host has sufficient CPU cores and RAM