There's something that does my voice even when I stay quiet.
I write a sentence, hit enter, and I hear it come back spoken by me. But it isn't me.
It's a model running on my server that learned how I sound.
What it does
I gave it a few minutes of my real voice. From there it caught the timbre, the rhythm, the way I say things.
Now I hand it a text and it reads it to me. In my voice. Without me having to switch on a microphone and redo the same line twenty times until it comes out right.
What it's for
The videos I make have a voice on top. Before, it was on me, every time: record, mess up, start over.
Now I write what it should say and it says it. If I change a word, I don't redo everything: I change the word and that's it.
It's not perfect. On some sentences it mishears, and then I rewrite them the way they should be pronounced. But it's fast, and it's here.
Why at home
There are online services that clone a voice better than this. I know.
But my voice is mine. I don't feel like uploading it to someone else's server so it can sit there forever.
It runs on my own machine, along with the rest. When I shut the server down, the voice goes quiet with it.
The point
I have a server that thinks, a memory that remembers, and now a voice that speaks.
Piece by piece, the room where I work looks more and more like someone. Except that someone is me, multiplied.
Alessio Cappuccio — xyz.vision