Speech Aid for People Losing the Ability to Speak

December 12, 2013 3:43 PM

Modeltalker has been around since at least the early 90s ... Modeltalker is a company that, for free, provides people with a synthetic version of their own voice and the software that lets users convert any text they want into that voice. It is continually updating it's software and in beta stages. But for people with onset neurological diseases that threaten to rob them of the ability to speak, Modeltalker will provide them with an 1800 word list to read. From that list, it will deliver a software program that contains their voice, the software and the tools to adjust the voice to make it as natural as possible. At some point, the company will make it product public. There are many synthetic voice programs, but only Modeltalker can make a synthetic voice out of your voice. For now, people can get a free version.
posted by CollectiveMind (13 comments total)

I have a relative that has never been able to speak. When we were kids she had a "word board." It said things like "Mom" "Dad" "Hungry" "Sad" "Angry." Then she got this speak & spell type thing that had a couple hundred words. Then she got an assistive device that allowed her to do sentences, but she was always upset that it only had a "boy voice" and that it sounded like the talking moose. It wasn't until the 90s when she finally got a girl talker.

It's difficult to understand how much lacking a voice affects your identity. These technology gains are great.
posted by cjorgensen at 3:52 PM on December 12, 2013 [7 favorites]

Is this the one that Roger Ebert used?
posted by Wild_Eep at 3:54 PM on December 12, 2013 [1 favorite]

Some of the samples I listened to on the site just sounded like garbled, chopped-up speech rather than anything clearly intelligible. 'Chee-ue-is s'unwich'.

I hope the technology for these things gets much better; it seems like speech synthesis was about this good ten years ago.
posted by pipeski at 3:54 PM on December 12, 2013

In the iOS world, see also All Access Menus (built by a friend of mine).
posted by Wild_Eep at 4:27 PM on December 12, 2013

Ebert used CereProc.
posted by RobotVoodooPower at 6:25 PM on December 12, 2013

Has anyone worked with FestVox, for comparison? Already public and released years ago and open source to boot.

The Festvox project aims to make the building of new synthetic voices more systemic and better documented, making it possible for anyone to build a new voice.

...

This work is firmly grounded within Edinburgh University's Festival Speech Synthesis System and Carnegie Mellon University's small footprint Flite synthesis engine.

This work has been supported be various groups including, Carnegie Mellon University, the US National Science Foundation (NSF), and US Defense Advanced Research Projects Agency (DARPA).

The claim that "only Modeltalker can make a synthetic voice out of your voice" seems unlikely to me, in 2013.
posted by XMLicious at 6:37 PM on December 12, 2013

Building a voice with Festival takes a really long time, and needs a very patient reader. I've used Flite extensively, though; its default voice is a bit robotic and decidedly Scottish, but is clear enough. It will run on quite small systems.
posted by scruss at 8:01 PM on December 12, 2013

After seeing what Ebert was able to do only because of his huge library of recorded speech, I would be willing to spend a few days reading a list of stock phrases & words into a laptop if it meant I could later do this.

I don't smoke, I don't chew tobacco, and I can't imagine what would rob me of my speech -- but I am a wicked motor mouth and this seems like a small price to pay for a big reward later.

Can I do this for myself and use it later with one of these packages?
posted by wenestvedt at 8:20 PM on December 12, 2013

Building a voice with Festival takes a really long time, and needs a very patient reader.

Well, if recording the 1800-word list described in the OP and then learning to use and applying the "tools to adjust the voice" goes quickly and doesn't require patience, these Modeltalker people might have a real breakthrough on their hands. If they've been around since the early nineties and still haven't released version 1 for public availability yet, maybe this is the speech synthesis equivalent of 20-year bourbon.
posted by XMLicious at 8:31 PM on December 12, 2013

Very cool
posted by ClaudiaCenter at 11:26 PM on December 12, 2013

After seeing what Ebert was able to do only because of his huge library of recorded speech

Is that right? I could swear I read somewhere that Ebert initially thought it would be no problem because of all those episodes of At The Movies and whatnot, but it turned out that there was always background music or other noise that would interfere with the process. If memory serves, the key turned out to be a commentary track for some DVD that was just him speaking with no other sounds.
posted by LastOfHisKind at 6:51 AM on December 13, 2013

"Only" was an overstatement. Thank you for mentioning other systems such as FestVox and Cereproc.
posted by CollectiveMind at 8:41 AM on December 13, 2013

Public speakers, podcasters and youtubers who recorded lots of material and without background noise, look out for this new type of spear phishing
posted by yoHighness at 1:45 PM on December 13, 2013