Vosk Speech Recognition

Installation

docker-compose up --build

The Vosk model is downloaded into /models while the image is built. To use a new model version in development, update VOSK_MODEL in compose.dev.yml, then rebuild the image.

Supported Audio

• Format: WAV
• Encoding: PCM-16 LE
• Depth: 24-bit
• Sample Rate: 16 kHz

Example

curl -X POST -F "audio=@example.wav" http://127.0.0.1:2700/recognize | jq