Project Polyglot

If you want to extract audio from youtube file,try this pafy module

making the prototype

1. download speech recognition module

git clone https://github.com/Uberi/speech_recognition.git
cd speech_recognition

2. install speech recognition module for python

sudo apt-get install python-pyaudio python3-pyaudio
sudo python setup.py install

3. test a transcribing example

cd examples
python wav_transcribe.py

output:

Google Speech Recognition thinks you said 5

4. INSTALL vlc for format converting (for manual conversion to wav file)

sudo apt-get update
sudo apt-get install vlc browser-plugin-vlc

5. for conversion using script

import subprocess
command = "ffmpeg -i C:/test.mp4 -ab 160k -ac 2 -ar 44100 -vn audio.wav"
subprocess.call(command, shell=True)

or just type the following in terminal

ffmpeg -i C:/test.mp4 -ab 160k -ac 2 -ar 44100 -vn audio.wav

Note: replace C:/test.mp4 with your file location

6. put window size of 5 sec and offset will be moving, which is like a moving window..

find most recurrent at a time and it considered winning word

7. language translate api, Now google has stop providing free translation API. So, pytranslate and goslate are not working.

https://translate.googleapis.com/translate_a/single?client=gtx&sl=en&tl=zh-cn&dt=t&q=hello%20how%20are%20you

decoding the url:
translate_a is the translate app
client is gtx may be google translator
sl is source language
tl is translated language
dt is the data representation
q is the data

8.speech synthesis hack

https://translate.googleapis.com/translate_tts?q=Hola%20ho&tl=es&client=gtx

decoding the url:
translte_tts is the text to speech method
q is the data
tl is the language translated
client is gtx may be google translator

9. after getting google converted voice(which is female voice), we don't want exactly those voice because the source voice is different.

So, we need to morph the speech you got in foreign language to the source language voice. http://www.cerlsoundgroup.org/Loris/

git hub:https://github.com/gesellkammer/loris14-extended.git https://www.clear.rice.edu/elec301/Projects02/lorisFor/loris.html

malcom research: https://engineering.purdue.edu/~malcolm/interval/1995-061/

report with matlab code

http://svr-www.eng.cam.ac.uk/~hy216/VoiceMorphingPrj

http://suendermann.com/su/pdf/gscl2013.pdf

kohlbrecher2013opensource.pdf

Heroku Deploying

need to have buildpack for scipy and numpy

https://github.com/thenovices/heroku-buildpack-scipy

then add

numpy==1.9.2 scipy==0.15.1

to your requirements.txt

add Procfile with line

web: gunicorn webclass.wsgi --log-file -

and my requirement file looks like this

Django==1.9.1
dj-database-url==0.3.0
dj-static==0.0.6
django-toolbelt==0.0.1
gunicorn==19.4.3
psycopg2==2.6.1
static3==0.6.1
whitenoise

SpeechRecognition==3.1.3
numpy==1.9.2
scipy==0.15.1