[DECtalk] DECtalk's development, and old speech synthesizer recordings

Thu Nov 3 17:41:45 EDT 2022

Hi,

I am just having some interesting speculations about DECtalk, that I thought
I would share with the list.

Have you ever heard Dennis Klatt's collection of speech synthesizer
recordings, that he maded in 1986?

He made a collection of 36 different recordings of various speech
synthesizers, detailing the history of speech synthesis, from 1939 to 1986,
includin early recordings of DECtalk, which was included in a paper in 1987,
accompanying an article called "Review of text-to-speech conversion for
English".

Here is a zip file with the entire collection:

https://www.dropbox.com/s/4vtuhnrt8xvllkc/Dennis%20Klatt%27s%20speech%20synt
hesizer%20recordings.zip?dl=1

The recordings are in au format, but I think most programs are able to play
them, otherwise let me know. I've included two different versions of the
recordings. Version 1 has narration before each recording, detailing which
synthesizer is recorded. Version 2 has no narration, but has slightly better
audio quality. I am quite sure though, that the narration is not Dennis
Klatt's own voice, since the article states the following:

"thanks also go to Kenneth Stevens, who served as narrator."

So, listen to recording 33, of Klattalk, which formed the basis of DECtalk,
and recording 35, which is of several of the DECtalk voices. Note that the
Paul from Klattalk, recording 33, and Paul from DECtalk, recording 35,
sounds almost, if not completely, identical, and also that they sound quite
like the later DECtalk versions, like 4.2 and 4.3, not identical, but quite
similar.

So, here I am asuming that the Klattalk recording is of the original system,
while the DECtalk recording is of the current DECtalk version, as of
november 1986.

But then, listen to the three recordings in the following zip file, which is
of DECtalk 1.8 and DECtalk 2.0, boath from 1984, and DECtalk 4.3 from 1996,
reading the exact same text:

https://www.dropbox.com/s/kbjrmoyhvig2of0/DECtalk%20Comparison.zip?dl=1

The thing that I find really interesting here is, that the Klattalk and
DECtalk recordings from Dennis Klatt's collection, to me atleast, sound much
closer to DECtalk version 4.3 than version 1.8 or 2.0. You can obviously
tell it's the same voice in all the recordings, but to me, there is a
difference boath in intonation and voice quality.

So, what does this mean? For me, it raises a few interesting questions:

1: What "DECtalk sound" came first?

If the Klattalk recording is of the original Klattalk system, then it must
mean that DEC took that technology, made the voice sound quite different,
and then changed it back later, all in a short period of time from about
1983 to 1986. Or, Dennis Klatt could just have used the current 1986 DECtalk
version to make the Klattalk recording, which could mean that the original
Klattalk, before DECtalk, sounded more like DECtalk 1.8 and 2.0. Another
reference could be recording 32 from Klatt's collection, the Speech Plus
Prose 2000 synthesizer. Speech Plus also licensed Klatt's work, like DEC
did, but developed it in a different direction. It had only one voice, which
was actually the voice Stephen Hawking used, and had a quite different
intonation than DECtalk. This voice, to me, sounds somewhere in between the
DECtalk versions, but possibly closer to DECtalk 1.8 and 2.0, which could
emply that Klatt just used the 1986 DECtalk for the Klattalk recording.

2: What did Dennis Klatt's own voice sound like, and how much was he
involved with DEC's development of DECtalk, after they licensed his work?
Did his voice sound closer to DECtalk 1.8 or 2.0, or to the Klattalk and
DECtalk recordings from his collection?

I haven't been able to find any recordings of Klatt's original voice on the
internet, so I think this is really interesting. Personally, I find it most
likely that his voice sounded more like the recordings in his collection,
because they, to me, sound more human and naturalt han DECtalk 1.8 and 2.0.
Also, that could explain why DECtalk 4.2, 4.3 and so on doesn't sound that
different from those recordings, because I suppose it would be difficult for
DEC to continue to develop DECtalk to sound as close to his voice as
possible, after his dead, unless they had several recordings of his own
voice, which, of course, is possible. Another thing to consider is the
"Doctor Dennis" voice from DECtalk, which Klatt's collection sadly hasn't
got a recording of. This voice was actually not available in DECtalk 1.8,
but was included in DECtalk 2.0, but sounded extremely different from the
later DECtalk versions. In DECtalk 2.0, the voice was very rough and the
breathiness was way louder than the actual voice, while the DECtalk 4.3
version sounds very natural, with a clear, understandable voice, and just
enough breathiness to make it sound nice, in my upinion. Here's a zip file
with a comparison of the two voices:

https://www.dropbox.com/s/igmp977sra5nom8/Doctor%20Dennis%20Comparison.zip?d
l=1

I believe I read somewhere, that the Doctor Dennis voice was supposed to
represent Klatt's voice in later years, due to his illness breaking his
voice, but then, it seems very strange that the later DECtalk versions would
make the voice sound much better than the DECtalk 2.0 version, because I
can't imagine his voice got that much better before his dead in 1988.

Best regards:

Aksel Christoffersen

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://bluegrasspals.com/pipermail/dectalk/attachments/20221103/d38ee45c/attachment.htm>