I want to get one thing out of the way first: My use of text-to-speech (TTS) for entertainment purposes is a joke compared to what it does for people with disabilities. Companies like CereProc do miracle work especially by giving young people with speech impairments fitting and even emotionally right voices.
I’ve been using text-to-speech for a long time now and watched it evolve from monotone synth voices like Microsoft Sam to the semi-real voices of Voicery, which sadly shut down last year.
That said, what is my deal? To me, TTS is like virtual instruments for text. Maybe I’m not a great pianist, but with these tools I can make my emotions be heard, in a cumbersome and weird, but also safe way.
Most of the systems I use (like Acapela Box) don't generate speech out of thin air, they work like samplers, with thousands of little recorded phrases that strung together make up words and sentences. Many of these these little microphrases contain interesting emotional residue. I’m mining for these traces of humanity.
But why go through all that mining trouble, why not write a script and then let professional voice actors do their work?
If TTS did not exist, I maybe would do that. But with TTS, I don't have to. It's a luxury. If I can write a verbal interaction between two people on my computer and listen to it, it will directly affect my game and tell me which situation could come up next.
I never (and nobody else could) finish a script and then plonk it into a text-to-speech converter, it would not work, the sentences would be too long, robotic, the artificiality not charming, but all too flat. In order for it to work, every sentence has to be brief, dependent clauses are rare. It shapes the writing, but in my opinion for the better. Every line is always written in conjunction with its virtual voice, working around its limits and strengths, word by word. It’s a slow process, like teaching a bird how to talk: Try this word, maybe it will sound better! Oh, that’s an interesting phrasing, keep that.
In a semi-funny game, timing is important and I like having it "down", not in text, but really fixed in a recording.
But don’t the limitations sometimes make you mad? Don’t you long for actual humans?
Usually, voice acting in games is recorded at the very end of a production, which means it can’t affect gameplay situations anymore. It is text brought to life, but too late for my tastes. As weird as that sounds, talking about computer generated speech, it would be less spontaneous. Yes, I do long for humans, but I’m also extremely impatient. If I had access to a professional studio and could direct the voice actors, I maybe would. But with my financial situation, I would have to resort to hiring people online, which would mean writing, sending it to a stranger, having them record it, then give feedback and then wait and wait some more. I WOULD love to try this, but then there’s also budget reasons, which I won’t go into much here.
So, TTS forever?
Actually, after my last game Beautycopter I thought I had finally exhausted this well. There are only so many interesting (often less zealous) digitized voices online. I made plans to finally move to the real world, with real voice talent. Then I started writing lines for Piggy Dreadful, my next project, and the familiar process was kicking back in, I found some digital voices I had not heard before... But like I said before, my budget is so flimsy, my choice is between silence, Animalese, or repeating everything from Japanese console games like Moon or Chibi-Robo!, but worse. Secretly, my plan this time is to use TTS for writing purposes, to create some kind of DEMO, that I can maybe hand over to voice talent to BRING IT TO LIFE. Piggy Dreadful will be more emotional than Beautycopter, it could very well be necessary.
Thank you for reading this! Here’s a trailer for my last game:
ellievsbear
blakekathryn
theartofmadeline
dededos
andulkaphoto
