BARF’s voice has bothered me for ages.
He’s gradually become much more than a way of delivering instructions in Galactic Gladiators. He complains. He panics. He’s sarcastic. Occasionally he’s even impressed by the player — although he’d probably rather not admit it.
The problem was that his voice never really reflected any of that.
Tonight I discovered Chatterbox, an open-source AI speech system capable of producing much more expressive speech than the text-to-speech I’ve been using.
So, naturally, I had to try it.
The first experiment was a familiar BARF line:
“Despite all expectations… you survived.”
And immediately there was something different.
It wasn’t just reading the words. The emphasis and timing actually changed the joke.
That led to experimenting with different levels of expression — from almost completely deadpan BARF through to considerably more alarmed BARF.
Then things got rather more interesting.
I created a reference voice specifically for the character and started testing whether the same voice could work in completely different situations.
A calm warning:
“Asteroids detected. Try not to hit the large rocky objects.”
A rather more urgent one:
“Incoming fire! Shields are failing. I strongly suggest you do something about it.”
And, following an unfortunate collision:
“Ouch! That hurts… you clumsy human idiot.”
That last one made me laugh, which is usually a pretty good sign.
The important discovery was that BARF could now remain recognisably the same character while changing his delivery depending on what was happening in the game.
Of course, repeatedly editing scripts to generate dialogue was never going to last long.
So BARF now has his own Voice Studio.
It lets me write a line, choose the type of delivery, adjust the amount of expression, generate alternative takes, listen to them and keep the performance that feels right.
Which is perhaps slightly excessive for a character whose main contribution so far has been offering advice, issuing warnings and occasionally insulting the player.
But BARF would expect nothing less.
The voice still needs refining, and plenty of the existing dialogue will eventually need revisiting.
But for the first time, I think BARF is beginning to sound like BARF.
And all of this started because I heard about an interesting voice synthesiser and thought:
I wonder what happens if I try that…