On-Premise-Speech-to-Text

A Flask API to convert speech to text using Offline Transcription methods - CMU Sphinx and DeepSpeech.

File Descriptions:

1. DeepSpeech.ipynb - Run this file to generate the DeepSpeech model and store the model files in a folder called 'deepspeech-0.6.0-models'. This step has to be completed before running anything else.

2. home.py - Main python file containing the flask APIs.

3. video_structuring.py - This python script converts the video/audio file into a .wav file (16kHz, 16 bit rate and 1 channel) of duration 50 seconds and saves it to the 'Files/Audio' folder.

4. cmu_sphinx.py - Python code to convert the wav file to text using CMU Sphinx.

5. deep_speech.py - Python code to convert the wav file to text using DeepSpeech.

Output:

Output will be stored in 'Files/Transcript/output.txt' file.

APIs:

Home.py contains two flask APIs -

1. /generate_transcript (POST)

Request body: (form-data format)
'method': 'cmu' (for cmu sphinx) or 'deepspeech' (for deepspeech),
'file': <The uploaded audio/video file>

2. /download_transcript (GET)

Sends the 'output.txt' file containing the transcript of the uploaded audio to the client.

Libraries:

Few of the libraries that have to be imported -

ffmpeg
pydub
flask-cors
srt
SpeechRecognition
download swigwin (and add to path)
download Visual studio C++ Build tools and add to path. https://visualstudio.microsoft.com/visual-cpp-build-tools/