How do I synthesize speech like in old video games?

December 26, 2005 5:47 PM

Wizard shot the food! What do I need to synthesize a voice similar to that in the videogame "Gauntlet?"

I'd like to create my own voice synthesis, similar to what you might hear in the old Odyssey 2 video game system, or the arcade game Gauntlet. Is it possible to do this without actually having the physical hardware from the arcade/video game system?

How are those voices created? Are they actual recordings that are then digitally converted, or is it an artificial computer voice, similar to the text-to-speech software currently available?

Extra credit: How would one create a "Cylon" voice off Battlestar Galactica, or the "Joshua" voice from Wargames?

posted by fatbobsmith to Computers & Internet (17 answers total)

Green elf needs food, baaadly!
posted by Brian James at 5:56 PM on December 26, 2005

"Paperboy" employed something similar to "Gauntlet"...
posted by johngoren at 5:57 PM on December 26, 2005

Best answer: The device is called a 'Vocorder' armed with that knowlege you will see them all over
posted by Mr T at 6:01 PM on December 26, 2005

...vocoder
posted by StickyCarpet at 6:16 PM on December 26, 2005

Although Mr. T is technically accurate, these days vocoder mostly refers to devices that will analyse and modify existing speech, and not devices that actually produce speech without a human source.

Most speech synthesis systems are referred to as text-to-speech (TTS) systems, and there's a number of them out there.
posted by Jairus at 6:25 PM on December 26, 2005

(...to clarify, most old very old video games used TTS.)
posted by Jairus at 6:26 PM on December 26, 2005

for kicks, try the third gauntlet link here! WOO! 5 seconds and 41 seconds into the clip: Elf needs food, badly!

and Jairus is 100% correct. everything back then used TTS vs vocoding.
posted by herrdoktor at 6:38 PM on December 26, 2005

Response by poster: Thank you for the clarification on the terminology. I've toyed with several of the TTS programs that Jairus linked to and for the most part they seem focused on more realistic TTS synthesis. I suppose I'm looking for something that will create that "retro" sound, from the classic arcade games.

As for vocoders, all the links I've found in google so far are for hardware devices. Do software-based vocoders exist?
posted by fatbobsmith at 6:54 PM on December 26, 2005

Response by poster: I guess I didn't search hard enough:
Cylonix - 18-channel software-based vocoder with Cylon preset
posted by fatbobsmith at 7:05 PM on December 26, 2005

I dunno, I've heard vocoders used in Kraftwerk albums and they don't resemble Atari arcade speech synthesis...
posted by johngoren at 7:25 PM on December 26, 2005

Best answer: According to the boyfriend, the voices in Gauntlet aren't actually synthesized, but are recorded and compressed using an algorithm called LPC (linear predictive coding). There's a special chip that does it on the Gauntlet arcade machines. You can record your own voice, then recompress it using an LPC converter like the one here.
posted by jenh at 7:30 PM on December 26, 2005

What you want is a Bitcrusher.

Something to crush the samples down to around 8 bit.

Soundtrack Pro has this...and therefore, I'd assume most of the audio processing software has that too. Acid...not sure about Audacity.
posted by filmgeek at 7:48 PM on December 26, 2005

Jenh's Boyfriend is right, they're recordings. There was an Amiga Power interview (early on, perhaps issues #3-5) with one of the creators.
posted by holloway at 7:52 PM on December 26, 2005

Response by poster: Jenh, that's exactly what I was looking for. A little google-fu got me to this piece of software, a real-time open source LPC encoder that, with a little tweaking nails the Gauntlet voice perfectly!
posted by fatbobsmith at 7:56 PM on December 26, 2005

Don't hurt the other players. Yet.
posted by stinkycheese at 9:52 AM on December 27, 2005

Er, shoot. Don't shoot the other players. Yet.
posted by stinkycheese at 9:54 AM on December 27, 2005

You could also grab a C64 emulator and some speech synthesis software.
posted by arto at 1:02 PM on December 27, 2005