Skip to content

Latest commit

 

History

History
108 lines (71 loc) · 1.91 KB

ReadMe.md

File metadata and controls

108 lines (71 loc) · 1.91 KB
Raw

Mimicry.ai

Copy the voice of anyone using machine learning.

Vagrant Dependencies

  • Install following dependencies
    sudo apt-get install -y git csh make g++ unzip tree
    sudo apt-get install -y libcurl4-openssl-dev
    sudo apt-get install -y python-virtualenv
    sudo apt-get install -y python-dev
  • Setup Virtual environment
    cd ~/mimicry.ai
    virtualenv venv
    source venv/bin/activate
  • Install Python dependencies
    pip install --upgrade pip
    pip install numpy scipy matplotlib ipython jupyter pandas sympy nose
    pip install theano lxml

    export CFLAGS=-I/home/vagrant/mimicry.ai/venv/lib/python2.7/site-packages/numpy/core/include
    pip install bandmat
  • Install dependencies
    cd ~/mimicry.ai/src/merlin/tools
    /bin/bash compile_tools.sh
  • Run test (Optional. Not needed)
    cd ~/mimicry.ai/src/merlin/test
    /bin/bash test_training.sh
  • Run demo
    cd ~/mimicry.ai
    source venv/bin/activate

    cd ~/mimicry.ai/src/merlin/egs/slt_arctic/s1
    ./run_demo.sh

Tasks

  1. Install dependencies

  2. Find voice of top actors, politicians

    • Rajini
    • Kamalkasan
    • Modi
    • Jayalalitha
    • P Sidambaram
    • Vadivelu
    • Karunanithi

  3. Train the machine with sample data

  4. Setup Sample UI

  5. Provide text for conversion

Audio download

https://pagalworld.me/ http://mp3lier.info/

Tools:

youtube-dl --extract-audio --audio-format mp3 https://www.youtube.com/watch?v=4CxkK1gam8o

Flask Server

  • Install dependencies

    cd ~/mimicry.ai source venv/bin/activate

    cd server pip install flask

Run Python server

python index.py

Request:

curl -XPOST http://api.mimicry.ai/mimic -F 'speaker=modi' -F 'content=I Love India'