KVR Audio

noouch · Post by **noouch** » Mon Jun 18, 2007 9:13 pm

Can anyone tell me how/with which tools this was synthesized? Sounds too fluid to be patched together from a phoneme bank (unless it's one friggin huge one), and unlike anything I've heard before, except maybe speak'n'spell and similar LPC speech synths...

http://noouch.de/audio/voxcollection.mp3

yonyz · Post by **yonyz** » Mon Jun 18, 2007 10:09 pm

maybe a Vocoder?
like Vokator or Orange Vocoder?

Modeler · Post by **Modeler** » Mon Jun 18, 2007 10:15 pm

Sounds like the Speech Synth in FL Studio on its "Munchkin" setting, sped up and fed through a phaser.

Well, it does!

gsoto · Post by **gsoto** » Mon Jun 18, 2007 10:23 pm

Doesn't sound synthesized to me. And too clean to be a vocoder, but I never got into vocoders, so I can't tell.

Modeler · Post by **Modeler** » Mon Jun 18, 2007 10:38 pm

gsoto wrote:Doesn't sound synthesized to me. And too clean to be a vocoder, but I never got into vocoders, so I can't tell.

Doesn't sound vocoded to me, the pitch varies too much (i.e. the carrier is too natural). Maybe it's just human speech that's been timestretched.

Anyway, I am the Gorfian consciousness.

EDIT: Reminds me of Software Speech - used on old Epyx / US Gold games on the C64. It was human speech, digitised at an extraordinarily low bitrate and compressed like hell. It wasn't synthesised but did sound very cool.

Modeler · Post by **Modeler** » Mon Jun 18, 2007 10:57 pm

I'll try again [wipes rabid foam from chin]...

I reckon this is synthesised, but there is a lot of distortion (intentional?). This is my quick and dirty attempt at a similar effect, could probably get it somewhere near with some tweaking. Amazed at how clearly it says "light bulb" and "wank".

Don't know what system it is, imagine it's some mid-80s Votrax hardware that pre-dates the system Stephen Hawking uses. Would be interesting to know though.

Scotty · Post by **Scotty** » Mon Jun 18, 2007 10:58 pm

Gorfian Robots .... a welcome ghost from my arcade past.

gsoto · Post by **gsoto** » Mon Jun 18, 2007 11:10 pm

lol, Modeler. Nice attempt.
But yours is clearly text to speech. Without being an expert I can hear the common artifacts of synthesized voice. In the original mp3 the articulation is too perfect, I can't spot a single strange thing in the entire audio. Listen to the s's, they are too natural.

Modeler · Post by **Modeler** » Tue Jun 19, 2007 7:32 am

gsoto wrote:In the original mp3 the articulation is too perfect, I can't spot a single strange thing in the entire audio. Listen to the s's, they are too natural.

Agreed, but it pronounces "reservoir" as "reserv-oyer".

noouch · Post by **noouch** » Tue Jun 19, 2007 8:04 pm

One thing I noticed is that all the S words start almost exactly the same. Modeler's example comes really close. Less distortion and breath and it's pretty much spot-on. Care to share the settings for this?

Modeler · Post by **Modeler** » Tue Jun 19, 2007 9:21 pm

noouch wrote:Modeler's example comes really close. Less distortion and breath and it's pretty much spot-on. Care to share the settings for this?

Sure, but I didn't spend much time refining it. Speech synth is "Large Male" set at pitch level C2, channel volume ~50%. Put it through Delay Bank with "Time" set to 1:00 and volume set at ~20%. I put the mix of that through the JCM900 simulator that comes with Guitar Suite (free) and put it on channel B - that's it. Childish swearing is optional.

Have fun.

Speech Synthesis