Skip to content

Latest commit

 

History

History
48 lines (43 loc) · 1.54 KB

README.md

File metadata and controls

48 lines (43 loc) · 1.54 KB

FakeYouHelper

Conversion and Speech-To-Text script to help with creation of FakeYou datasets

What does this do?

  1. Resamples all wave files to 22050Hz 16-bit mono (FakeYou standard)
  2. Transcribes text from them and generates a list to use for AI training

Setup

You must have Python 3 and Git

Clone this repository

git clone https://github.com/cavv4/FakeYouHelper.git

Change directory

cd FakeYouHelper

Install requirements

pip3 install -r requirements.txt

Install ffmpeg

Linux:

sudo apt install ffmpeg

Windows:

  1. Download build from here
  2. Open the archive and extract the contents of ffmpeg-[version]-essentials_build in C:\ffmpeg\
  3. Go to your Computer properties and open Advanced system settings
  4. Click on Environment Variables
  5. Find the PATH (or Path) variable in the bottom window, select it and click on Edit...
  6. Click on New and paste in C:\ffmpeg\bin
  7. Click OK and you are done

Download a VOSK model for your language from here
and extract its content into a model folder like this:
VOSK model folder structure

Usage

Put all your wave files into an input folder
Wave files folder structure
Start the script

python3 fakeyou.py

That's it, now wait