correlator/deepspeech-rest-api

REST api for mozilla deepspeech voice recognition engine

deepspeech rest api

Simple api over mozilla deepspeech voice recognition engine.

Endpoints

POST /api/v1/stt - Just look at curl command below.
Speech data may be provided in whatever audio format which ffmpeg is able to convert to wav,
so you probably don't have to worry about this at all.

$ curl -X POST -F "speech=@speech.mp3" http://127.0.0.1:8000/api/v1/stt 
{"text":"experience proves this","time":0.9638644850056153}

Setup

0. Checkout repository

git clone git@github.com:zelo/deepspeech-rest-api.git

1. Download deepspeech data model

1. Look at requirements.pip file and find what is the current deepspeech library version used by this api
2. Go to https://github.com/mozilla/DeepSpeech/releases and find release doc for this version. It should contain link to download data model. For version 0.7.1 it's https://github.com/mozilla/DeepSpeech/releases/download/v0.7.1/deepspeech-0.7.1-models.pbmm
3. Download and extract model from above package to <repository_root>/model.pbmm

2.1 Using docker

1. Enter <repository_root>

2.1.1 Using docker-compose

1. Run docker-compose up

2.1.2 Using docker run

1. Build image docker build . --tag zelo/deepspeech-rest-api:0.7.1
2. Run docker run --rm --publish=127.0.0.1:8000:8000 --volume=$(pwd)/model.pbmm:/app/model.pbmm:ro zelo/deepspeech-rest-api:0.7.1

2.2 Running outside of docker

Just look at the content of Dockerfile it contains complete instruction to setup app under debian