Project Diplomatist

What is Diplomatist ?

Diplomatist is an auto audio transcribe(audio2text) and translate(lang2lang) tool.

APIs

• 0 - CMU Sphinx
• 1 - Google Cloud
• 2 - Bing API
• 3 - Houndify API
• 4 - Baidu DeepSpeech (macOS/Linux Only)
• 5 - Azure Speech Service (WIP)

Key Features

• Audio Transcription
• Text Translation
• Loopback Capture and Microphone Record
• Output as SRT

Support Format (Transcribe from File)

Actually, Diplomatist only support to transcribe from .wav file and .aif file. For other formats, it will try to convert from other format to WAVE which was supported by pydub, and the FFmpeg is required in your environment.

You can get FFmpeg from here: https://www.ffmpeg.org

Requirements

• python 2.x/3.x
• pip install -r requirements.txt
• pip install -r requirements[.osx|.linux].txt (for macOS/Linux)

Before Use

config.json

Most of required configurations will be loaded from config.json, there are some default values already, if you want to change these configs please update config.json.

• API: name, cred, model, alphabet, etc.
• LOOPBACK_CAPTURE: location of LoopbackCapture
• SRT: save output string to srt file
• RECORD: location to save temp record files
• PROXY: http/https proxy servers

LoopbackCapture

In Windows, compile LoopbackCapture under /LoopbackCapture and set the path of LoopbackCapture.exe as Loopback_Capture_Path in diplomatist.py.

/LoopbackCapture folder included 3 solutions of win32, one console application written by C++, one console application written by C# (recommended) and another one is a .dll library written by C#.

In macOS, the mac/LoopbackCapture.py support to capture sounds from output device but not support to capture Loopback from device directly. The workaround is to route what is playing on the computer digitally back to the input without using a cable. Suggest to use Soundflower.

In Linux, the linux/LoopbackCapture.py support to capture sounds from system but it require a tool avconv which is a part from "libav-tools" package to support to record audio use command line with specific format specific duration and from specific device.

Detail: LoopbackCapture README.md

Credentials

Most of APIs required a credential to use the API. You can register a account in the website and get the credential file or key. Most of them have some free quota each month.

• Google Cloud API: https://cloud.google.com/speech/
• Bing API: https://azure.microsoft.com/en-ca/pricing/details/cognitive-services/speech-api/
• Houndify API: https://www.houndify.com

Usage

> python diplomatist.py -s 15000 -a 1 -l en-US -t zh -o sub.srt

Usage: diplomatist.py [options]

Options:
-h, --help            show this help message and exit
-m, --mic             record sounds from microphone
-f AUDIO_FILE, --file=AUDIO_FILE
                        audio file which to be transcribed and translated
-s TIME_SLICE, --slice=TIME_SLICE
                        time slice of each wave file
-a API, --api=API     0 - CMU Sphinx, 1 - Google Cloud, 2 - Bing API, 3 -
                        Houndify API, 4 - Baidu DeepSpeech
-l LANGUAGE, --lan=LANGUAGE
                        language which to be transcribed
-t TRANSLATE, --tran=TRANSLATE
                        translate to another language
--qt                  runs UI with QT
--tk                  runs UI with Tk

UI

Currently, Diplomatist has a very simple UI based on Tkinter or PyQt4. You can use it through ui.py with arguments --tk or --qt, other arguments are same as diplomatist.py:

> python ui.py -s 15000 -a 1 -l en-US -t zh -o sub.srt --tk

# or 

> python ui.py -s 15000 -a 1 -l en-US -t zh -o sub.srt --qt

Before you use ui.py with --tk, please install python-tk package:

In Linux:

1. > sudo apt-get install python-tk

Before you use ui.py with --qt, please make sure you have installed PyQt4/PyQt5 in your environment:

In Windows:

1. Get PyQt4/5-***.whl from http://www.lfd.uci.edu/~gohlke/pythonlibs/#pyqt4
2. > pip install PyQt4/5-***.whl

In macOS:

1. > brew install PyQt

In Linux:

1. > sudo apt-get install python-qt4