Slashdot Mirror
Bell Labs, Preserving Delicate Sensibilities
LuserOnFire writes: "There is a PigDog article talking about the Bell Labs Text-to-Speech Synthesis. The amazing thing is not the technology itself, but that fact that Bell-Labs has a checkbox next to it that says 'If you plan to enter text which our system might consider to be obscene, check here to certify that you are old enough to hear the resulting output.'?!?! Like if you are old enough to spell a swear, you don't know what the word sounds like?" More fun than a TI-99/4A with speech-synthesis card. Those wouldn't say the bad words at all.
Much better sounding text-to-speech from ATT
If you had the single purpose "Speech Synth" cartridge, it would refuse to say any word not in its very limited vocabulary (it would map words you typed to a phonetic spelling).
There was also the "Terminal Emulator II" cart that would allow you to say anything from a BASIC program. The quality of speech was much lower, since there was no phonetic mapping going on.
It wasn't that TI censored the speech synth cart, they just didn't include a phonetic version for swear words.
...speech-synthesis (and I mean it, no digitalized vocabulary or the like, no expansion modules, it was part of the OS) was implemented as a filesystem.
So you could do
CLI:> copy textfile SPEAK:or
CLI:> type textfile >SPEAK:or use the SAY command
CLI:> say -p120 "Hello, old sucker !"You would do that in 1986 [and it would sound a lot worse ;-)]
The Amiga sythesized all on the fly.
However, I found that Bell Stuff nice. It is possible to hear the human voice behind that. I trapped myself whishing this would have gone, so to hear a native computer voice while sounding like a human rather than a robot.
Seems I got the Digitanoia...
Gee, I still love my Amiga
But you had to get the right cartridge - that "speech editor" cartridge that came with the Speech Synthsizer module only had a limited vocabulary.
What you really needed was the cartridge "Terminal Emulator II"; then it was possible to write programs in TI basic that just fed the speech synthesizer bits of phonetics out of which you got speech. Actually, the phonetics were fed to the TE2 cartridge, which handed you back a huge list of numbers to feed to the synthesizer, and then your program had to feed the result to the synthesizer. I suppose that it would be possible to take the list of numbers, save them off somewhere, and then have basic programs that could swear even without the TE2 cartridge, but that never came up. Really, getting the TI to say "you asshole" once was enough.
The cover of the TE2 book even showed a little caption ballon next to the computer that read "I can say anything". And it could, sort of, in that stilted way that it had.
The TI/99-4a speech synthesizer could produce "obscene" words. They weren't in the default dictionaries though. You had to learn about phonetics and piece the sounds together yourself until it sounded just right.
Ahh, the memories.
--
if you type obscenities into their system, that's your problem. But if they send obscenities back at you, they should be sure that you've explicitly authorised it.
as of 13:30 :) good job all :)
errr....umm...*whooosh* *whoosh* Is this thing on ?
Enter, listen, and end the debate.:-)
The best warning ever, must be this:
Don't use nuclear weapons to troubleshoot faults.
It comes from SAFETY RULES FOR US STRATEGIC BOMBERS2 C3AF4F2snlbxq'|dc
--
echo '[q]sa[ln0=aln80~Psnlbx]16isb15CB32EF3AF9C0E5D727
echo '[q]sa[ln0=aln80~Psnlbx]16isb572CCB9AE9DB03273snlbxq' |dc
Either that or it's my employer's LOVELY internet connection again.
Jon Acheson
All opinions expressed herein are my own, and not those of my employers, who are appalled.
The reference is wrong. According to this page, both British and Australian computers had been used for playing music at least six years previously.
The music of CSIRAC, the first Australian computer, has recently been recreated using an emulator and rebuilt hardware, and a CD has been released.
Go you big red fire engine!
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
but Lucent's page wasn't able to utter a phase through my web browser.
All my dialog boxes eventually catch my attention by speaking the message if I'm not paying attention. Even more important with OS X since I can be concentrating on some other task.
Text to speech is cute, not very difficult and not computationally demanding.
Speech to text is a very different kettle of fish.
MSBPodcast.com The opinions expressed here are my own. If you don't like 'em... Think up your own stuff.
Quick! Someone pipe the DeCSS source through there so the RIAA can go after Bell Labs!
grubTrolling is a art,
Actually using the terminal emulator 2 cartridge you could use the system to get the speech module to say anything you wanted it to say (sometimes requiring creative spelling).
You could always OCR it, and get all the profanity with none of that annoying music getting in the way :)
Caution: contents may be quarrelsome and meticulous!
Your right to not believe: Americans United for Separation of Church and
I thought MS Bookshelf has only been out for the past three or so years
MS Bookshelf was one of the first "shovelware" CD apps out there. Back when buying a single-speed CD-rom (with controller card!) for $900 was a good deal...
---------------------------------------------
Recursive: Adj. See Recursive.
"Pardon me, Stewardess? I speak Jive."
-- Barbara Billingsley
--
--
"Outlook not so good." That magic 8-ball knows everything! I'll ask about Exchange Server next.
Wait, why can't you just read through the lyrics? It's much easier to search text than audio.
And if your post was meant as a joke, well, it wasn't that funny.
i remember when i was 14 or so, and i discovered that ms bookshelf had an audio file for the sample pronunciation of "motherfucker" i thought it was absolutely hilarious to have that monotone voice demonstrate to me how to swear.
ah, the memories...
#define F(x) int main(){printf(#x,10,#x);}
F(#define F(x) int main(){printf(#x,10,#x);}%cF(%s))
First of all, I'm surprised no one has yet mentioned the old program "DR. SBAITSO" (Sound Blaster Artificial Inteligent something or other) that came along with really old SoundBlaster sound cards (it came with my 8-bit SoundBlaster PRO). The voice sounds almost the same as this bell labs program, so I'm assuming it used the same algorithm. Anyone know where to download Dr. SBAITSO?
On to how text-to-voice works. This was an excellent situation for using an artificial neural network. I don't recall the exact topology of the net, but the lower (input) neurons contained different letter combinations often found in the english language. The output neurons were, obviously, the sound that needed to be produced.
Neural Networks, being the excellent statistical analysis tools they are, would train on many words and determine their "error" by comparing the network output to the actual pronunciation of said words. The network would then adjust the weights through many iterations to reduce this error to zero. Then, new words are fed to the input neurons, and then we have the voice you hear on the Bell Lab site.
This technique for text-to-voice lent a lot of weight to neural network programming. The people demonstrating the algorithm were particularly clever, in that as the neural network was training (i.e. adjusting internal weight values to reduce error to zero), they would test different words on the network in a child's voice. This basically gave the effect of sounding like a child slowly learning to speak. One can imagine how impressive this would be for someone who doesn't really know how the algorithm works.
Of course, in the old days, we used to get our yuks by programming the TI (or other prudish voice synthesizers) to say things like:
Fa Queue Ice hole
Talk sec stew me beach
etc. Needless to say, I was much younger in those days. Today I'd rather have them say things like "Thanks for calling RetroAxle Inc. For a list of the ways technology has made your life better, please press 1..."
bukra fil mish mish
-
Monitor the Web, or Track your site!
Eloi, Eloi, lema sabachtani?
www.fogbound.net
The Trash 80 software usually said "I cannot say that", but for a**h***, it would say "Rectum", which, given its low fidelity, was far more amusing sounding.
You are in a maze of twisty little passages, all alike.
This stuff sounds only marginally better than speech card add-on a college friend had for his TRS-80 and that was in 1982.
:-P
It, too, filtered profanities, but foh-net-ik spelling solved that problem.
That's bugged me for years. In Kubrick's 2001, they could talk to HAL and HAL would respond in a pleasant human sounding voice. Okay, it didn't do a great job singing, but not only do we in the real 2001 not have a computer you can converse with (Eliza notwithstanding), but we can't even do the speech! Of course, Clarke never envisioned marvels like http://www.amiallyourbaseornot.com, so I guess we have other advantages in the real world vs. science fiction.
You are in a maze of twisty little passages, all alike.
I especially enjoyed using this site to make very humourous new mail alerts. My personal favorote was to use the german one to say thing in english with a german accent. Hours of fun!
spoonz
---
The checkbox exists only in the english version. The foreign language versions don't have it.
Interesting, what that says about the maturity of english speaking people.
Confused
...how bad you think this thing is, perhaps you
should take a look at some of the examples to see
just what it is capable of. This think is not the
mere equivalent of something you ran on a TI-99 or
a MAC once, it's really quite a sophisticated thing.
mefus
--
um, er... eh -- *click*
mefus
In Open Society, GPL Software frees YOU!
ISTR that was one of the Flanders kids that said that.. in the episode where Moe opened a family restaurant.
Here is a link to a reference: The first time a computer played music was in 1957, at Bell Labs in the United States. The song was called Daisy, which is the same piece that the intelligent computer HAL (in Stanley Kubrick's film version of Arthur C. Clarke's science fiction novel 2001) starts humming as it is being disassembled. Naturally, this is not a coincidence, but rather the intention of the director to return the computer to its "childhood state" (in a double sense) as it loses its advanced electronic identity
http://cm.bell-labs.com/cm/cs/who/dmr/odd.html
Ralph: "My freakin ears!"
My favorite is all the conservative websites out there that designate movies to be "good" and "bad" simply by counting all the times they have the word fuck....
just a little fucking rediculous if you ask me...
Doug
Venn ist das nurnstuck git und Slotermeyer? Ya! Beigerhund das oder die Flipperwaldt gersput!
Woops.. I think you're right...
They're the same voice actor, and I was
just going off of what I remembered
(I have a very audio oriented memory).
Thanks for the catch...
Doug
Venn ist das nurnstuck git und Slotermeyer? Ya! Beigerhund das oder die Flipperwaldt gersput!
I remember visiting that exact same web site about three years ago. As nonsensical as that obscenity disclamier may be, it's not really news anymore. Bell Labs is just covering their butts.
What I'm curious to know is how much their Text-to-Speech synthesis has improved in the past few years. Heck, even the local Weather Channel has more realistic synthesis than this.
Made the rounds a decade ago, from a radio 4 interview honoring 40 years since the battle of britian. Many modified versions of this story have circled the internet in humour files and joke lists.
I've heard the original audio on a BBC request show. Quite hilarious when you hear the whole thing. This is the closest I've found in a web search to what I remember the interview was like.
BBC INTERVIEWS AN RAF PILOT...
Gerry Wills, the famous BBC commentator, was interviewing Gerherd "Zibby" Zebdrehah the equally famous Polish WWI fighter ace who flew for the British. The interview went like this...
BBC: So please tell us Captain Zebdrehah about your most intimidating foe from those years.
Zebdrehah: I remember being jumped by 4 or 5 Fokkers. My God, the sky was thick with those Fokkers and every where I turned they were on me instantly...
BBC: I should just inform the radio audience that Captain Zebrehah is talking about the aircraft his opponent usually flew, the "Fokker" fighter plane.
Zebdrehah: Ya, maybe usually but these Fokkers were all flying Messerschmits!
the AC
Hemos is like...sci-fi fans;he thinks technology is cool, but he hasn't bothered to understand the science it's based on
Somewhere I still have the Usenet article where somebody was trying to get an Apache server to "speak Java" (some JVM problem I think), and was told that he had forgotten to signal with SIGHUP after making his changes. Except that U is right next to I on a qwerty keyboard, and so the poster mispelled it SIGHIP.
One of the followups said, no, if you send it a SIGHUP it'll speak Java, but if you send it a SIGHIP it'll start speaking Jive, and then proceeded to list a bunch of HTTP response codes all in "jive".
The only one I remember is "404 that file is NOT in da house."
You cannot apply a technological solution to a sociological problem. (Edwards' Law)
In high school, I did some work for a local TV station, and I still remember filming an interview with a "hometown celebrity" who flew in some of the older warplanes.
Fokker (sp?) makes a lot of warplanes. Fokker also sounds a bit, um, obscene when spoken by somebody with a gutteral voice.
"...an' then we hit some ack-ack from the border guards an' that ol' Fokker started twisting so bad..."
Had to do quite a bit of trimming on that interview before running it.
You cannot apply a technological solution to a sociological problem. (Edwards' Law)
With TI Basic and the Terminal Emulator II cartridge, you could make the Speech Synthesizer say anything you want.
It was more fun for me to type in this small program from MICROpendium for TI Extended basic, which made the Speech Synthesizer make all sorts of strange-ass sounds. I believe one game I had for it (in Extended Basic) made a frog croaking sound from the routine.
I stil have no idea how they got the female voices from the game Parsec in it though. Looknig from an old TI magazine, they DID digitze a woman's voice to do it.
Ahh, the speech synthesizer, an underexploited piece of hardware for the TI, as well as the MBX expansion system which had speech recognition!
Man, I can remember my mom getting pissed that they had it in the movie and it got the PG rating. Yeah, my virgin ears.
They're putting dimes in the hole in my head to see the change in me.
I am flabbergasted that computer text to speech still sounds like a "drunken swede in a garbage can". Would someone please explain this to me? Why is it so incredibly difficult to synthesize a human voice that sounds even remotely intelligible?
It seems as if this technology is following a strange inverse of Moore's law where it gets better by only half as much each year.
Are you sure there was a "Speech Synth" cart? I had tons of TI-99/4A junk (some to be sold on Ebay), and never heard of it. I think you mean the "Extended Basic" cart. I had a later version of it from MircoPal, and it had a small list of words that could be used with the command CALL SAY("HELLO") and so on. If there was such a cart, it must have been a piece of junk.
As other people said, it was the TE2 Cart that provided good text-to-speech. Someone else posted that you had get/feed numbers for that to work with TE2. I think he is confusing it with the "Extended Basic" cart. There was a hackish workaround of getting small samples of speech published in some magazine (Home Computer Mag I think), but was a bitch to use. The TE2 cart let you do this in it's modified version of BASIC(from my memory):
10 open #1,speech (I think thats what you use)
20 print #1, "Fart Fart Yams Hobo in my room"
Or it was something simular to that. It's text to speech was VERY good for the time, I only needed to modify some words to better speech rarely.
you could also modify the speech tone/speed by doing:
30 print #1,"//30 59"
or something close. If you used values outside valid ranges (It never checked), you could get some crazy hissing and growling sounds.
It was the easy version of basic that the TI had that got me interested in programming. God Bless 'em.
The folks at Bell Labs have a sense of humor, anyway.
On their pre-generated samples page, the English sample is a computer-sung rendition of "Bicycle Built for Two." This is the song which the murderous HAL 9000 in Clark/Kubrick's 2001: A Space Odyssey sang as he was being put to sleep.
[
If schoolkids had such a fun program as this, they'd never go to the school library to look up naughty words in the dictionary.
was it saying "help, ive been slashdoted!"
Interestingly, their voicemail product (which sucks, btw) presumably uses the same engine, and passes any word you want.
actually I guess that's not all that interesting.
Just remember to check the box on your scanner software that says your old enough to see offensive words. ;)
The best of these would be capalert.com. He hasn't rated "Freddy Got Fingered", but he has nothin' good to say about the South Park movie. "*South Park* is another movie straight from the smoking pits of Hell."
Fade To Black somehow got Cap from CapAlert to do a Q&A column. It's a hoot.
--
... it will give the kiddies the idea if they haven't thought of it already! :-P
http://www.commondreams.com/
Hey Dennis Miller called he wants his bad stand up back.
"`Ford, you're turning into a penguin. Stop it.'" -THHGTTG
The Festival speech synthesis system page also has a web-based text-to-speech converter here, without the filter. It's free software, and does a pretty good job.
;-)
The automatic voice pitch is pretty neat; I built a hardware text-to-speech converter around 10 years ago, and it only produced a monotone voice that got pretty annoying after a while. Don't feed Festival raw HTML documents, though - it can cause the voice to get deeper and deeper until it has to reset the pitch.
But I was also really disappointed with the results. It doesn't sound much better than the old Apple II setup my chem teacher would wheel into class when his voice went.
I blame cheap sampling over the past decade or so. The ability to record a voice actor's voice and use that has precluded any real advancement on the synthesis front.
People talked about Jar Jar as the first virtual main character (and he wasn't even first) but they seem to forget-- his voice was nothing near virtual.
For better or worse, it seems like it will be a numberof years before we have an artificially generated voice as annoying as Jar Jar's.
--
SO YOU'RE GOING TO DIE: The Comic for Dealing with Death
Damn, I miss the good ol' days sometimes...
...the amazing thing is that if this feature weren't included, somebody would sue when their eight-year-old is exposed to synthetically-spoken profanity.
My mom is not a Karma whore!
Great, now we slashdotted free speech.
Monkey sense
I tried to make that thing work and it kept failing. anyway...
The fact that we are to a point where we have to declare that we are old enough to hear what we typed, things have gone way too far.
Now I don't wan to get off on a rant here, but it seems as technology increases, so does the desire of everyone to make sure they have no reason to get sued. Personally I don't think the two are related. I think it has more to do with the fact that we have constantly been making more and more laws without enforcing the laws in existance. This leads to more and more lawyers there to interpret and fight the laws while at the same time, twist the laws so that they can make an optimum dollar. Combine that with the fact that most parents in today's society never take responsibility for their children while expecting the greater public to do so and you have a manifistaion such as the one written about here at Bell Labs. If we expect these things to change, we need to take responsibility for ourselves and our loved ones and stop putting the burden on the schools and the rest of society. After all, I want to be able to control what my under-aged children see/read/whatever, not those who think they know what's best for me and mine. But as usual...
Of course, that's just my opinion, I could be wrong.
because if it were objectionable then the sytem wouldn't play it unless you checked the box first! but you're still right, typing it into the textbox would take a lot longer than just reading through the lyrics.
"Ask me about Loom"
This stuff sounds only marginally better than speech card add-on a college friend had for his TRS-80 and that was in 1982.
It, too, filtered profanities, but foh-net-ik spelling solved that problem.
We've got one of those at home. To clarify, I believe it's the software that does the profanity filtering rather that the speech synthesiser device itself.
On another note, I seem to remember an old game called "688 Attack Sub" having an amusing profanity filter. You could send messages when playing the game multi-player, but it would alter any of the swear-words it knew about - so while you might type out something rather threatening to your opponent, it would come out as "I'm going to kick your donkey, Mother Theresa!!", or something similar :-)
I had one, and you didn't miss much. We had one game that used the speech synthesis. "Alpiner". The speech thing made fun of you when fell down. Said stuff like "Smooth move, sport".
With decades of research, exponential increases in computational output and DSP, when do ya suppose voice synthesys will sound like a voice?
(Of course, all I get is silence, and a 500 error.)
"Old enough to hear obscenities"? How old is that, anyway? I know kids aren't allowed near rated R movies, but you can already hear "fuck" (The Seventh Day) and "shit" (Star Trek Generations) in PG-13 movies. I'm also guessing you can curse in public, even when kids are around. Yes, I know about that One Case where a guy was busted for swearing in front of a kid, but that's one case in one state.
The earlier the better, in my opinion. There's nothing wrong with a little harsh language.
Speak and Spell Simulator ... that thing was ripe for abuse in the wrong hands =P
E.
Build Your Own PVR/HTPC news, reviews, &
Actually, Timothy got it wrong-
The ti99/4a WOULD say the bad words, and it wasn't a speech synthesis card, it was a module that plugged into a port on the right side of the keyboard/console.
I liked my ti99/4a, my mom ran a ti education center, and had a network of 14 of them with the BIG expansion cases.
A host is a host from coast to coast, but no one uses a host that's close
Ok idiot, read more closely before you post shi-at like this. The article says text to speech, not the other way around. It would have been funnier if you had said something about them using alt.cowboyneal.fanclub as a test subject.
The Blaster Master Fighting for Truth, Justice, and Evil Pie since 1979
you could actually run song lyrics off a CD booklet into this device and determine whether or not the CD deserves the advisory label without having to listen to the songs themselves. Next up: preventing misuse of your X-Ray vision
Bryguy
microsoftword.mp3 - it doesn't care that they're not words...
I used to entertain myself by making Speak-n-Spell say dirty words all the time. The problem with using a computer to do convert text to speech is there are bastards out there that think some words should not be heard, and a computer is smart enough to try to filter those words. Fortunately even a small child is a lot smarter than a computer.
Mom wouldn't let me have the speech card. Said that HAL was bad enough and she didn't want to hear things like: "You bugger gorillas" or "Wax on; Wax off" all day long.
Comparing it to Windows will be a moot point, since El Dorado is going to have a 40% larger code base than XP.
I dunno -- remember Jamie's article about the kids who screwed up the output of fortune to HTML and put the words "I put the shotgun in an Adidas bag and padded" on their web site? Where Jamie was irate that the police asked them some questions before realizing it was a misunderstanding and dropping the matter? Where Jamie and one boy's father were claiming that the kids now have a "police record" (absolutely false, according to the usual understanding of that term)?
In the course of showing what obviously angelic youngsters these are, Jamie writes:
..whose domain name contained the school's name and the Fword. This is a word, by the way, which G. obviously typed in to register the domain but which he was too polite to use over the phone. By the time we hung up, he had me embarrassed for saying it.
Uh, yeah. The owner of fuckwestbeverlyhigh.com is a candidate to replace Miss Manners.
Unsettling MOTD at my ISP.
"I call it a Hawking chamber."
Eerie.
.. doesnt work.. "Internat Server Error" /.-effect?
Some
ha det godt,
Platy
Tongue-tied and twisted, just an earthbound misfit, I.
I'm sure in a parallel universe somewhere, that checkbox was not placed on the page. The page was then encountered by that same idiot woman who spilled McDonald's coffee on herself. She was offended that her computer said "ass" to her after she typed it in, and promptly sued Lucent/Bell Labs out of existence. And probably whoever made her computer, too.
Come to think of it, I'm surprised I haven't heard from her attorneys yet since I still rip her publicly for being a litigious moron every chance I get.
~Philly
I AM a mirror running. Mail me your text and I will call you back by phone to read your text aloud.
I can do the male voice, the female voice, in French, Dutch and English.
More fun than a TI-99/4A with speech-synthesis card. Those wouldn't say the bad words at all.
Yeah, the TI-99/4A voice synth was quite limited, but we ended up with...
Anybody want a peanut?
Anyways, the thing is that it's like TV. They don't want it that something this widely used might have obscenities. There is enough parents that would raise hell over something as small as that, and likely, they just don't want to deal with it. That's all.
01101001 01100001 01101101 01101110 01101111 01110100 01100001 01101100 01100001 01110111 01111001 01100101 01110010
Does this mean the already brokenup bell labs will be broken down into Bell labs and Bell MP3 distribution?
Whoa...
Dancin Santa
...of a speech synthesizer that has no problem at all with swearing and obscenities is Stephen Hawkings' speech synthesizer. He seems to have a second career as a gangsta rapper. Listen to his mp3s and be amazed...
It seems to be missing the "Drunken Hobo" and "Jive" voice option. Welcome t' da damn Bell Labs text-t'-speech system. Step off, Pharoah.
He doesn't know how to use the seashells?!?
Instead of what you said, you'll get a receipt printed out
Or:
**computer voice** you are fined one-half credit for violation of the verbal parameter
As in "Demolition Man."
Along with a receipt of demerit from M$, which would be a fun alternative to toilet paper!
My Karma was at 49, then they switched to words. All that work for nothing!
If, like me, you prefer the Stephen Hawking's kind of voice, go get IBM's ViaVoice TTS SDK (free with registration). Yup, they have a Linux version. It's cool, software-only, fast, understandable and will utter any obscenity you care to throw at it. -- Danny C.
Microsoft does have this already! They recently licensed Speechify, which is based on AT&T Watson (which is Belllab's engine).. so, here come's SAPI 5 to the rescue.
Many moons ago, when I was a leet Commodore 64 Basic programmer who also used hex editors to disable copy protection in such games as "Raid on Bungeling Bay". Me and my friends had software that would take a string of english words and turn them them into audible words. We would use this to call up friends and insult them using the words of the C-64 through the telephone. Great fun was had by all, Even more fun would have been had If "bad words" were enabled. I can see Bell engineers keeping themselves amused for hours with this thing. =]