Dockerfile for speech to text (STT) based on Kaldi ASR (http://kaldi-asr.org/) Kaldi's 'chain' models (type of DNN-HMM model) used Model released by api.ai team (https://github.com/api-ai/api-ai-english-asr-model)
In the command line run following commands
docker build -t stt .
docker run -it stt
file in.wav will be processed and output should contain following text:
/opt/in/in.wav HELLO THIS IS SPEECH TO TEXT RECOGNITION FOR JOKER PROJECT
output also contains a lot of debug messages (ignore it).
Record your file first with following command (works under Linux)
mkdir in
arecord -d 10 -fS16_LE -r 16000 -c 1 in/in.wav
or you can prepare in.wav in any other software. File format: 16 bit, mono 16000 Hz
now you can run docker container:
docker run -it -v `pwd`/in:/opt/in stt
(c) Abylay Ospan [email protected], 2017 https://jokersys.com