mxaviersmp/voice-api

API to allow STT and TTS

Voice API

This is the API responsible for STT and TTS

The endpoins are:

/: Health check.
    request:
        GET /

/stt: Processes and audio file and outputs its transcription
    request:
        POST /stt
        {
            'audio': '<wav audio file in base64 format>'
        }
    response:
        {
            'text': '<audio transcription>'
        }

/tts: Synthesizes a text sentence into an audio file
    request:
        POST /tts
        {
            'tts': '<text to be synthesized>'
        }
    response:
        {
            'audio': '<wav audio file in base64 format>'
        }

Environment variables

• A list of keys for the wit.ai api: `WIT_KEY=["<key 1>", "<key 2>"]
• A google drive link with the model: `MODEL_LINK=

You can run this as a docker container

• docker build --tag <image-name> .
• docker run -p 5025:5025 -e WEB_CONCURRENCY=<number-of-workers> <image-name>
• If you want to update the model, add the UPDATE_MODEL=1 variable

To run locally:

• clone the TTS repository with git clone https://github.com/flych3r/TTS -b pt-br
• install the api requirements: pip install -r requirements.txt.
• run: bash start.sh

In the voice interface, add VOICE_URL=http://localhost:5025 to the .env file.