Audio-stories

Example Stories

Format as provided by the design tool

For a story that starts with a classifier, see data/samples/audio_story_classifier_start.json

For a story that starts with an audio file, see data/samples/audio_story_audio_start.json

Format as accepted by the front-end

Story that starts with a classifier:

{
    "nodes": {
        "0": {"type": "classifier", "labels": {"indoors": 1, "outdoors": 2}, "thresholds": {"indoors": 0.5,"outdoors":0.5}},
        "1": {"type": "audio", "next": 3, "audio_file": "audio/indoors.mp3"},
        "2": {"type": "audio", "next": 3, "audio_file": "audio/outdoors.mp3"},
        "3": {"type": "audio", "next": null, "audio_file": "audio/changes.mp3"}
    }
}

Story that starts with an audio file:

{
    "nodes": {
        "0": {"type": "audio", "next": 1, "audio_file": "audio/changes.mp3"}
        "1": {"type": "classifier", "labels": {"indoors": 2, "outdoors": 3}, "thresholds": {"indoors": 0.5,"outdoors":0.5}},
        "2": {"type": "audio", "next": 4, "audio_file": "audio/indoors.mp3"},
        "3": {"type": "audio", "next": 4, "audio_file": "audio/outdoors.mp3"},
        "4": {"type": "audio", "next": null, "audio_file": "audio/changes.mp3"}
    }
}

Server

A Flask server provides access to the CLIP model.

Setup

Make a conda environment and install the dependencies.

conda create --name "audio-stories" python=3.8
conda activate audio-stories
pip install -r server/requirements.txt

Running the server

Here is the command to run the Flask server:

./server/run.sh

This script automatically attempts to set Google TTS credentials, then starts the server.

Run ./server/run.sh --help to see optional args.

Running the server tests

There are unit tests for the audio saving/loading functionality and integration tests for the API. Run them like this:

python3 -m pytest server/test

Files

server/index.py: code to run the server, very bare-bones
server/create_app.py: sets up the app, includes API endpoints
server/model_utils.py: code for using the CLIP model
server/save_audio_files.py: code for storing and loading audio stories on the server
server/speech_generator.py: code for doing text-to-speech
server/test: tests

mjoerke/audio-stories