So this week I’ve been sick at home and spent my time tinkering on something that I’ve been thinking for a while.
This is looking more and more like something out of a Mr.Bean episode. So I thought I’d share it for all of you, just for giggles.![]()
My goal is to have a natural sounding TTS, that can read texts for me (or rather with me), as it would help my semi-autistic brain to concentrate.
So first, a bit of research that escalates to looking into offline LLM assistants. A few days of playing around with Gpt4all, Jan, and a bit of Newelle. Got a finnish made Poro-34B to work at the very maximum limits of what my PC can handle. Ok, cool, but this was not what I set out to do, it can’t make audio…
Back to researching and found out about Chatterbox-TTS. Looks perfect: does what i need and claims to have finnish language capabilities.
Did not find a desktop or CLI app that would utilize it directly, so now I have to do the thing, I try to avoid: Build from GIT!
As someone with zero coding skills, this hardly ever manifests in anything good.
There are a few projects, that have done what I need. Downside being, I personally dislike the web-page approach as a GUI. As a KDE user, would prefer a Qt interface, or even a cool old-school ncurses one… But hey, close enough!
I clone the first repo. Learn how to create a Python virtual environment, made sure I have all the dependencies installed and start the process according to the git instructions. It fails - have to install python-devel. Try again, it fails - search the web - have to install protobuf-devel and protobuf-compiler. Try again, it fails - search the web - From the Onnx-package git issues, I got the feeling that Python 3.13 might be too new. Rage quit.
Installed Debian 13 to a virtual machine. A different repo, with a project that does the same thing, comes first at the search results. Cloned it, compiled it. It runs, but is unable to generate anything and I don’t know where to look for error messages. Rage quit.
Look at another repo on Fedora. It has a docker version available. Downloaded Docker-desktop as instructed. It won’t install, gives an unsolvable dependency error. Rage quit.
Back to the Python. Download sources for python 3.11. Compile it with instructions from the repo. → make altinstall. Somehow it works and didn’t break anything!
Try compiling the chatterbox git with older python. Pip is unable to download any packages due to missing SSL-certifigates. Rage quit.
Learn about Conda. Watch a few videos and read three instruction pages and still misinterpret the usage, as the demonstrators talk about .yml -files and I have no glue what tools are needed in the environment. Ask from chatgpt, it gives an answer that works! Made an Conda environment with Python 3.11. Lost track what was the git project I had previously tried, so I cloned one, witch turned out to be yet another one. It did compile in the conda environment, but running it shows errors at code lines. Rage quit.
Now I try to calm my self down and find an organized way to focus on one of the projects. This is one of the rare occasions that I’d like to have web page history available.
First step: choose and pin one to the bookmarks toolbar… Or maybe I could just try to pip install chatterbox, as referenced in their own documentation. Here we go again, anyone with adhd knows the feeling when your mind giving different solutions to the same problem, at the same time… Wish me luck!![]()