People get strongly attached to speech synthesizers, even (especially?) the old robotic-sounding ones, whether they're using it to speak for them (as with Hawking) or read to them (e.g. blind people). I know a blind guy who was so attached to the ETI-Eloquence synthesizer (developed in the 90s) that, to use it on his Mac, he did a hack involving the WINE Win32 compatibility layer. Some blind people favor the more robotic-sounding synths in general because those tend to be better at high speeds than the more natural-sounding voices based on real recordings.
The lesson is that software companies shouldn't be quick to kill an old product or let it die just because they have a newer one that they think is better.
Edit: Perhaps the next generation of blind techies will become attached to something open-source, like espeak-ng (which ships with the NVDA screen reader for Windows). Then, if they want it frozen in time at version X, that will be feasible.
Interesting. Hawking appears to have preferred his old CallText for similar reasons to why I prefer chunky old bitmap fonts in my terminal windows rather than something like Consolas: One part nostalgia, and one part noise-robustness.
> The lesson is that software companies shouldn't be quick to kill an old product or let it die just because they have a newer one that they think is better.
I wish Microsoft had stopped trying to "improve" Windows after Windows 95, UI-wise. Maybe port it to the NT kernel and offer support for modern hardware like GPUs and USB, but that's it. Too many of their improvements to the experience aren't, really.
I‘d be delighted with a Windows 10 kernel and technology stack sporting a Windows 2000 GUI. That represents the nostalgic apex for me. XP’s default ”Fisher Price” skin was when things turned sour.
> software companies shouldn't be quick to kill an old product or let it die just because they have a newer one that they think is better
OMG, so very much this!
The problem is that there was a very brief period of time in the early days of the development of the computer (by which I mean like 1975-1995 or so) when just about everything new was an improvement. That implanted the idea in people's heads that NEW==GOOD because it was (mostly) true for about 20 years. But nowadays technology has matured and it is no longer invariably true. I have a five-year-old MacBook pro running Mavericks that I much prefer to anything on the market today.
I wonder if one of the reasons Hawking was so tied to that specific voice and hardware was so that there could be a way to verify that a recording was actually him. If nobody else had that hardware and software combo, it wouldn't be possible for anyone else to put words into his mouth.
With that in mind, it was nice to read that the original CallText boards as well as the emulator are with the family. That should provide the security to ensure that nobody can impersonate him in the future.
He was so tied to that specific voice, because that was his voice. If someone told you that today, you get a completely new voice, would you? It just so happens that this particular synthetic voice was attached to an extremely popular person, and it was his identifying trait.
Lots of people do. Politicians are well known for changing their speech later in life in order to convey a different feeling, usually "confident and statesman-like".
Personally I modify my speech by context, just as I modify my writing. When I'm "home" (where I grew up, my parent's house) I emphasise my accent a little more, just as I use local slang.
Some people, subconsciously echo a speaking-partners accent - changing the tone, timbre, inflections, and such between conversations.
But yes, you're right that voice was strongly linked to his identity as is often the case for prolific public speakers.
Hmm, I'd recognise the change in slang as form of code-switching, but the change in tone/accent I didn't think was part of that specifically as it's within the same language.
If using a different accent (different range of frequency, different speed, different tone) isn't changing one's voice then I'm not sure what really counts? At this point it sounds like you're trying to use a non-physics based concept for voice [the parameters of sound produced by vocalisation].
The movie "Do I Sound Gay" has an interesting take on this kind of thing. There's a stereotype of the "gay voice" (a bit higher-pitch, with sibilance on the "s"es, commonly incorrectly called a lisp). The 'intensity' of the voice-modifications depends largely on one's social environment. I don't remember the movie using the term "code-switching", but it certainly did seem like the same thing. I notice this myself - if I'm with other gay men I do usually adopt a bit of a changed speech-pattern totally unconsciously.
Speech is so nuanced; it really seems true that 80% of what you say has little to do with the words themselves.
English word choice is confusing here because voice can be "the style of speech/writing".
By changing [style of] their speech I mean the frequency, timbre, intonations, accent, nasality, etc., and NOT words, or content (which obviously can also be changed).
There are often key elements that match based on morphology of the vocal apparatus, but if I speak like a toff with a cold vs. a barrel-chested Nor'n Irelander there's little ID left.
Good mimics, those who have radio/TV shows for example, can if they choose leave nothing for a normal listener to distinguish their speech from that of the subject of their mimicry. Impressionists tend to charicateur their subjects somewhat, however.
In UK it used to be relatively common to have a "telephone choice" that would sometimes confuse even close acquaintances in to thinking you were someone else.
Another common example is when taking different languages one has a different voice.
Can't it? I have no idea but if you can train to speak in a different voice, how is that not changing your voice? Maybe I won't be able to speak like someone of the opposite gender because I don't have vocal chords like that, but learning to speak like someone with a similar body should be doable, no? Or can speech recognition systems really pick this out somehow and is there really the concept of a unique voice to a person?
Actually, though, the voice was a standard, default part of the DECTalk system, called "Perfect Paul". There were a few other options. https://m.youtube.com/watch?v=8pewe2gPDk4
In general, there's no security to be had in unique the sound of someone's voice. It is too easy to impersonate.
> it wouldn't be possible for anyone else to put words into his mouth.
You can copy normal voices, you can copy his voice and you can fully impersonate people and even create fake videos.
If security was his intent, he should have used cryptographic signatures for his texts or voice recordings, but even that can be compromised.
I think he just liked his voice and wanted it to be unique. It's understandable and most people would do the same.
I know of one other commercial synthesizer based on the Klatt model, DECtalk. (In fact, DECtalk is the product that Dennis Klatt went on to create after his research at MIT.) I wonder if DECtalk is close enough that it would fool most people. Here's a clip I made with DECtalk 4.50 (the last good version of that synth IMO, from the 90s):
The CallText 5010 is a descendent of DECtalk, so the voice he used has been around. Hawking's voice was based on Perfect Paul, which the DECtalk Express I have here is set to, and sounds very much like him. You can still buy these devices and a software version is available.
Just curious, do you still use a DECtalk Express day to day, or do you just have one for the sake of nostalgia? What do you use a speech synthesizer for (e.g. screen reader)?
It's just for fun. I got it because it sounds like Hawking. It came DOA from a music shop and it just turned out the battery pack needed to be serviced. Anything with a serial port can use it to speak.
I understand that his voice was his identity but our voices naturally and distinctly change as we mature and age. I am left to wonder why it was so important that his voice remain frozen in time?
Also, though it borders on surveillance, this text-to-speech technology offered the byproduct to record every spoken word that Professor Hawking said in lecture and made available as searchable text. He had authored every conversation he had. Does such a repository exist in his estate?
I would imagine part of it is that our voices change gradually. Even in extreme cases, like when we go hoarse or get sick, there's an essential progression to link our new voice to the old, and there's an identifiable common thread. Meanwhile, for Hawking, it would have been a sudden, disjoint change - and with the exception of the mentioned upgraded CallText voice, any replacement would have been completely new, utterly unrelated to his prior voice; Imagine how disconcerting it would be to awake with a stranger's voice, with no expectation of getting your own back.
Exactly. And whereas biological human voices deteriorate with time, the proposed replacements for Hawking's voice were supposedly improvements (though not necessarily improvements to him).
This really hit me when I watched The Force Awakens. I was so weirded out when I realized I was expecting Han and Leia to sound the exact same as they did in Return of the Jedi. The difference in their voices is really stark. I was much better prepared for hearing Luke's voice in The Last Jedi after that.
I knew they had gotten older, I knew their voices were different, but it never hit me until I saw them playing Han and Leia again.
The lesson is that software companies shouldn't be quick to kill an old product or let it die just because they have a newer one that they think is better.
Edit: Perhaps the next generation of blind techies will become attached to something open-source, like espeak-ng (which ships with the NVDA screen reader for Windows). Then, if they want it frozen in time at version X, that will be feasible.