This article is based on this Youtube video from Aitrepreneur
While you can find some decent Text-to-Speech tools online, like:
https://cloud.google.com/text-to-speech
https://www.naturalreaders.com/
It is harder to find free solutions to do text-to-speech on your computer like bark, to generate voice audios. I don't recommend to use this github to install Bark, because it can be tedious, especially if you already have many "pip install" solutions related to pytorch on your computer. I had for example an incompatibility that created this message: "No GPU being used. Careful, Inference might be extremely slow!" and indeed without a compatible graphic card with the correct setup of pytorch (see this "issue"), Bark is not going to do text-to-speech very well on your computer.
To install bark, i suggest you to go to this github that provides an Automatic1111 style web UI
Find the link to the "one click installer" (direct link), and follow the instructions. This will install Bark in a conda style virtual environment with all the requirements that are needed.
If you feel limited by the maximum length of 15 seconds of the wav files, go to this other github and download the github of Bark Infinity. To prevent a similar type of error ( "No GPU being used. Careful, Inference might be extremely slow!") and to not reinstall Bark, on Windows, use "Bark infinity" this way :
- (The Bark WebUI has to be installed)
- (Download a copy of the "Bark Infinity" github and unzip it in a directory on your computer)
- Within a Windows command line (cmd), go to the directory where "bark webui" is installed and Run "run.bat":
This will start the virtual environment and the webui.
- Do a control-c to exit the webui server
- In your command line window, go to the directory where "Bark Infinity" is installed and more specifically to the directory of the file "bark_perform.py"
- (do: pip install soundfile
if you didn't do it before: it helped me to solve a bug for some reasons)
- Use the text-to-speech script.
Example of command:
python bark_perform.py --text_prompt "It is a mistake to think you can solve any major problems just with potatoes... or can you? (and the next page, and the next page...)" --split_by_words 35
Finally, if you are unhappy with the quality of your wav, you can enhance them with tools like this one from Adobe.