Background
As I often do, I read books to learn and improve myself. I also enjoy teaching and helping others when I can, so this is something I try to do regularly.
However, as time goes by (and so do the years) I have slowly come to realize it is becoming increasingly harder not only to maintain focus when reading/studying a new topic, I also forget things faster.
It is not an uncommon occurrence to re-read an “old” article or book, and to become surprised with things I had learned before. I have come to the inevitable conclusion this will only happen more frequently with time.
Although I do enjoy reading and learning, if I keep reading and the learning the same things over and over again, I will not evolve. Eventually I will gradually lose my ability to help others in topics of relevance, something I cannot live with.
So I decided to take a new approach. Re-learning old concepts is now something I have accepted, but I do get to choose how I re-learn them. I have decided to make a series of instructional videos (10-15) minutes long, freely available on youtube, free lectures for all to see.
Not only does this achieve my end goal of helping others more efficiently, it is also a quick and convenient way of re-learning things for myself. It also allows me to add my opinions, experiences and additional context, which is something I value immensely.
Problem
I have several scripts done, and I have power-point like presentations done as well for them. The issue now is that in order to make videos, I need to record sound. I have it from several sources, that my voice is … not ideal. Not only that, should my initiative succeed and my channel gather traction, I will be barred from certain jobs in the country I live because it is a common occurrence for managers and bosses to “avoid hiring known people”, for company political reasons.
In order to protect myself and to create videos that are somewhat acceptable I have decided to use text-to-speech audio generation.
However, all the tools I have found so far are in Python. Many of them are abandoned and even when they are not, there are always dependency issues with the likes of PyTorch and Tensor and so on.
I have not yet seen a Python project where things just work because there is always a dependency issue somewhere. Even the creators of some of those tools acknowledge this. To make things even more complex, some of these tools require specifically NVIDIA GPUs, which I don’t have.
Question
So I was wondering if there are out there any Elixir tts generation tools that one could use. Because Elixir uses BEAM, I assume requiring an NVIDIA GPU is not a necessity. Furthermore, I also assume I wouldn’t have to constantly fight with Python dependencies.
Are there any tools in Elixir that follow this requirements? Or is it too early in the Elixir ecosystem for such tools to exist yet?
