Conversion and Speech-To-Text script to help with creation of FakeYou datasets
- Resamples all wave files to 22050Hz 16-bit mono (FakeYou standard)
- Transcribes text from them and generates a list to use for AI training
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:
- Download build from here
- Open the archive and extract the contents of
ffmpeg-[version]-essentials_build
inC:\ffmpeg\
- Go to your
Computer properties
and openAdvanced system settings
- Click on
Environment Variables
- Find the
PATH
(orPath
) variable in the bottom window, select it and click onEdit...
- Click on
New
and paste inC:\ffmpeg\bin
- 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:
Put all your wave files into an input
folder
Start the script
python3 fakeyou.py
That's it, now wait