Chip Directory
Mirror sites
Mailing list
Have this site translated
 Datasheet Archive
 Embedded Systems
Host site

Sound chips

Voice synthesizers

Every x months I get a question about the (very old) xxx256-AL2 voice synthesizer from GI. It was sold by Tandy once. Here is some data about it and about other voice synthesizer chips.

What we are talking about

256 CTS256A-AL2 text to speech controller $5 (GI)
256 SPO256-AL2 voice synthesizer (GI)

More info

With Altavista:

Other voice synthesizers

OKI voice synthesizers are: But they are also already quite a few years old...

Hualon and Philips also make voice synthesizers.

Date:         19960724
From:         Neil Robertson
Organization: MSE Ltd
Newsgroups:   comp.arch.embedded,alt.comp.hardware.homedesigned,alt.comp.hardware.homebuilt
Subject:      Re: A request for help (and don't laugh!)

We make a range of boards which replay EPROM* stored speech messages which get used for lift announcements, amusement machines and museum exhibits and amongst many other things.

The board which may interest you has one 64K EPROM*, runs from +9V, drives a small speaker and starts with a dry contact.

I'm not sure what the .WAV file format is, we use 8 bit PCM* but I guess you could convert a .WAV file.

A tribble wouldn't be the strangest application we've had but its close!

Not sure what the price is. E-Mail me if you're interested and I'll send you a data sheet.

Neil Robertson <>

Development Engineer
Morgan-Smith Electronics Ltd

tel: +44-1707-264583
fax: +44-1707-265898

From:    Jaap van Ganswijk
Subject: Re: Voice Chip

Hi Barry,

On 19961010, you wrote:

I am looking for a voice chip that can reproduce with good fidelity a celebrity voice in a 10-60 second prerecorded message. The chip would be inserted in a promotional publication. Perhaps you know of companies that specialize in this type of chip?

Perhaps the next fragment can help you:

ISD makes a chip that allows voice, music, or other sounds to be recorded into a greeting card or other small product. We make chips with record/playback times of from 10 seconds to 2 minutes currently and will introduce a device that records as long as 4 minutes late this year.

For instant information, you can look at our WWW page at

This web page includes all our data sheets and applications material in PDF format. Additionally there is an on-line product guide that is a quick reference to all our current products.

You can also look in the Chipdir on the functional page under: Audio, sound, speech, voice.

Firms like Oki, Philips and the DSP* makers do things with sound among others.

And don't forget to use a WWW/newsgroup search engine like AltaVista or Deja!

Date:         19960913
From:         Bob Langlois
Organization: Hitachi Data Systems
Newsgroups:   comp.robotics.misc
Subject:      Re: Digitalker Chip Set?

Bryan Rittmeyer wrote:

I saw a chip set called the "Digitalker" in the latest Jameco Catalog.

If this Digitalker is the same as the one I know of then its a National Semi chipset from the mid 70's. Its basically a set of roms with digitised voice which feeds the decoder and amplifier chip. It has some small sentances but mostly Phonems (i think thats how its spelt.) I used to have one wired up to a trash 80 thats how old it is.

Perhaps you would be better looking at some of the new eeproms that store up to 8 bits of data in a cell. Check in the chip directory and you will find it or look back a couple of months in this newsgroup Im sure that someone was talking about it.
I just found the article. I must have saved it.
That sounds similar to the "Voice Record/Playback IC*" available at Radio Shack (catalogue #276-1325). It has a 20-second capacity, and can store more than one message (voice, music, sound effects, or tones). It includes application notes, example schematics, and PCB artwork tips. Price: $24.99 Canadian (about $18 USA) dollars.
The 28-pin chip is an ISD1000A, from Information Storage Devices.

Regards, Bob Langlois
Perth Western Australia.
Date:         19991009
From:         Randy Carlstrom
Organization: RC Systems, Inc.
Subject:      Another voice chip

In case you'd like to update your database of voice chips, we recently introduced a chip set. If you're interested, you can see the chips at

The RC8650 voice synthesizer chip set is the latest creation from RC Systems. Our 16+ years experience in providing leading edge text to speech solutions has been integrated into silicon, now available in an easy to use chip set. You'll be amazed at the amount of functionality and versatility that we've built into these devices: unlimited vocabulary, on-the-fly voice control, tone generation, recorded voice playback, addressable voice channels, multiple input and output protocols... the list goes on.
Thought that a PC was the only way to generate speech from text? The RC8650 enables you to add text to speech capability to virtually any embedded system, quickly and painlessly, without a PC and virtually no overhead. You can send text messages to the RC8650 via an asynchronous serial port, printer port, or SRAM-like microprocessor interface. The audio output is delivered as either an analog signal or digital audio, if you need to do further processing of the audio stream. All the RC8650 asks for is 17 mA of operating current (5 V) to keep it happy, or only 2 uA when you're not using it. Think a PC would ask so little?

See RC Systems for address data etc.

Date:    19961208
From:    Dale Wellborn
To:      PIC Mailing List
Subject: [PICS] Voice Synthesis

Do any amongst you have experiences (good and bad) about integrating a speech synthesizer (with a custom vocabulary) into a project/product?

I'm contemplating such an endeavor, but have many questions. 8-}

My expected volume is quite low (<10000), and am unaware of what's available (in terms of development tools, etc) for small users like myself.

TI has a new family of devices denoted: MSP50P3X which look interesting.

And I'm unaware who else makes such beasties.

Any and all comments are appreciated.

At the moment I'm somewhat "net-challenged" 8-} ; are you aware of a more appropriate "voice-synthesis" mailing-list/news-group?

Date:         19961208
From:         Greg Keeney
Organization: Altair IV Productions
To:           PIC Mailing List
Subject:      Re: [PICS] Voice Synthesis

A cheap way of doing it would be to store your voice in a pulse width form, and then play the sample back out through a smoothing capcitor, which would integrate the pulses back into sinusoidal(ish) patterns. This is a old game programmers hack from back in the C-64 days!

Date:    19961208
From:    John E. Nelson
To:      PIC Mailing List
Subject: Re: [PICS] Voice Synthesis

Try the newsgroup comp.speech They have postings on speech generation (hardware and software) as well as voice recognition.

Date:    19961209
From:    Dave Cox
To:      PIC Mailing List
Subject: RE: [PICS] Voice Synthesis

I used a chip years ago by Votrax, called the SC-01A which is a phonetic voice synthesizer. I think that these are the same chips that were used in the speak-n-spell. There are basically 64 phonemes (speech primitives) and 16 tones which you can construct any English word from.
Not the easiest thing to use, but it does the job.

Date:         19961209
From:         Mark R. Dzmura
Organization: Phalanx Micro* Technology
To:           PIC Mailing List
Subject:      Re: [PICS] Voice Synthesis

Dave Cox wrote:

I used a chip years ago by Votrax, called the SC-01A which is a phonetic voice synthesizer. I think that these are the same chips that were used in the speak-n-spell.

Not so. Texas instruments used its own home-grown technology for the Speak-and-Spell and other similar talking toys. The technology was based on a mathematical technique known as linear predictive coding, or LPC, wherein a time-varying set of coefficients are used to parameterize a simple model of the human vocal tract. This technology is one of the precursors of low-bit-rate vocoding techniques in vogue (voc?) today.

The VOTRAX technology is a different beast. It was originally designed by the Votrax division of Federal Screw Works (believe it or not!), which had military applications of the technology in mind.

Regarding techniques for getting sound out of a PIC: the big problem is the tradeoff between quality, horsepower and bit rate. The more horsepower you have, the higher the quality audio you can record and/or playback at a given bit rate (due to the more sophisticated compression algorithms you can use).

As fast as it is, the PIC just bogs down at multiplication and division of multiple-precision numbers.

If you have the EPROM* space, you can run straight PCM* out a DAC at the bit rate of your choice; I have done this with 5x-series parts and external PAL*'s to drive the address lines of the EPROM*'s.

(This points out another problem with the PIC - address space!
Not to denigrate the PIC, though; as we all know, it is perfect for so MANY applications!)

Another possibility is some of the interesting chips available from OKI semiconductor and Yamaha. Both companies have big product lines featuring audio A/D and D/A technology. In particular, OKI's voiceband-quality chips using ADPCM* (adaptive differential pulse code modulation) offer central-office grade telephone quality at 4 k bytes per second. (This from 12-bit linear samples taken at a base sample rate of 8 KHz.)

A few of OKI's chips are perfect for PIC applications: they include a simple state machine, address generator, and address lines for static or dynamic RAM or EPROM*. The controlling processor need only twiddle a few lines on the chip to record/playback. With an 8-bit data port you can even force-feed the part for tone generation and such.

(I think that part is the MSM5832; all of these are in the MSM-family. Ask your OKI rep for a copy of the OKI Voice Synthesis Data Book).

Of course, if you have the horsepower, more efficient compression schemes are yours for the asking: The European GSM standard, used for digital voice in just about everybody's digital cell phones except ours (US), offers near-telco quality at 1.65 k bytes per second. This is about 100 k bytes per minute, very reasonable considering the alternatives we've come to expect!

Plan to have a 386 or 68k around to do the crunching, though... The problem is that algorithms like GSM are fairly computationally intensive and require 24-bit or greater precision in the math.
My hat is off to the implementor of GSM for the PIC16xx.

Many companies needing high-quality text-to-speech are using a software-based technology such as SmoothTalker from First Byte corporation. (I first used this technology a decade ago, so it's been around for a long time.)

Also available at the high end of speech synthesis are such hardware/DSP-based solutions as DecTalk (originally from Digital Equipment Corporation), and the Prose products from SpeechPlus (formerly Telesensory Systems); though these are fairly pricey solutions.

Finally, if you're talking about high-end audio, such as CD/DAT-quality stereo, plan on using a DSP* (Motorola 56K, Analog Devices, TI, etc.) and A/D - D/A's from such high-end suppliers as Burr-Brown, Analog Devices, and Crystal Semiconductor. Their Delta-Sigma conversion technology is fairly easy to work with and can offer studio quality.

And to think I've totally ignored speech recognition...

Date:    19970628
From:    Jan Zumwalt
To:      Multiple recipients of list CHIPDIR-L
Subject: SP0256

I am looking for source of the ASP256 voice chip or a close replacement.
Any ideas?

Date:    19970629
From:    Michael A. Tyborski
To:      Multiple recipients of list CHIPDIR-L
Subject: Re: SP0256

Many of us moved from the SP0256 to the RC Systems V8600 speech modules.
This system sounds better, has power management, and built-in text-to-speech. If you don't mind spending $150 (qty 1), I know you'll love the product.

Date:    19970630
From:    Robert L. Smith
To:      Multiple recipients of list CHIPDIR-L
Subject: RE: SPO256

On 19970629 Bas van Rossem wrote:
Do you mean the SPO 256 from General Instruments?
It has become obsolete many years ago.

Do you happen to know of a replacement? Any allophone talking device.

Check out the Voice Recording and Playback ICs, by ISD.

They employ analog mode EEPROMs and can store an amazing amount of
voice in an single IC*. These are true "systems on a chip" and
feature up to 4 minutes of duration.

2045 Hamilton Avenue
San Jose, CA 95125

Date:    19970727
From:    Luis Miguel Brugarolas
To:      Multiple recipients of list CHIPDIR-L
Subject: RE: SPO256

I am looking for alternatives to the SPO256-AL2 speech chip. Is this a "sound" chip or a "speech" chip?

This chip synthesizes 'alophones' somethink like 'speech primitives'. I used it ten years ago, and you can obtain with simple text to alophones conversion programs (I did an easy one for spanish), something you can understand when heared many times. It helps to know previously what the chip says :-).

I think this may be completelly obsolete now!

Date:    19971229
From:    Sam Haile
To:      Multiple recipients of list CHIPDIR-L
Subject: speech IC*

I am interested to build a Talking clock (part of my project at college). I know how to build the basic digital clock but not sure about the talking part of the clock.
Can any one help? Please recomand any IC* you may be aware of.

Date:    19971230
From:    Jeremy N.E. Proffitt
To:      Multiple recipients of list CHIPDIR-L
Subject: Re: speech IC*

Try using a DAC/ADC pair. Record your voice in one, and play back with the other. Use a computer to capture and burn EPROM* or FLASH and add, well heck:

Input lines to select
the number to say.

            ----------|             |
            ----------|   memory    |
            ----------|             |
                      |             |   Outputs to DAC
         |------|     |             |     8 bit is all you need
         |      |-----|             |
         |      |-----|             |
         |------|     |-------------|

Get my drift???

Anyways - bg micro* use to have a serial port text to speach card resonably cheap. Check them out. I have this card and it's a good one.

BG Micro also has a cool set of sound chipset chips. Check them out. Only $10 for their sound chip set.

I have used the parallel port for data capture in the past, so if you need any help just hollar. All of my code though is in assembly for a pc. Can help you with any other language, but my best field is assembly.

Date:    19971230
From:    Ivan Zilic
To:      Multiple recipients of list CHIPDIR-L
Subject: Re: speech IC*

Take a look at
They produce interesting ICs (ISD1420, 2500, ...) which could be used in
your project.

Date:    19971230
From:    Bhamidipati
To:      Multiple recipients of list CHIPDIR-L
Subject: Re. speech IC*

In fact you can use low-cost speech synthesizers from OKI semiconductors. They have a flexible range of these IC*'s or you can use very simple to use chips from ISD.

Date:    19971230
From:    António José Gomes
To:      Multiple recipients of list CHIPDIR-L
Subject: Re: speech IC*

One idea that comes to my mind is to use a microprocessor (I think that with carefull programming a simple microcontroller will do) and digital samples of the numbers and necessary words.
Although I've never tryied that my self, I think that with proper care the comutation from one sample to another would not introduce a transient sound capable of distorting the final sound and with this approach you can use the voice you like, even yours!
I believe that to this kind of task you can use 8-10 bits to encode the sample (256 to 1024 levels) and a sample would be about a few K long (it's an estimated guess, considering a sample ferquency of about 8 KHz).
If you would like to discuss this idea a little further, feel free to email me any time!

Date:    19971231
From:    Bhamidipati
To:      Multiple recipients of list CHIPDIR-L
Subject: RE: Re. speech IC*

Sam wrote:

Thanks for the reply and also the information you send me. Are you aware of OKI or ISD does accept external clock signal?

I presume that by external clock signal you mean to say sampling clock. Yes with Oki it is possible & well with ISD chips I have to recheck. I have recently come accross a project titled "talking doorbell" in elektor electronics which uses Eletech VP2500 recorder/playback chip. If you are not able to locate this article I can mail you.
I hope you have access to web and in that case you can get connected to the Chipdir or Oki site and can have access to data sheets as well as application examples. But even the design using ISD chips and ofcourse using VP2500 is very small and its matter of time you can really come out with a good design.

Date:    19971231
From:    Someone from Xtremetech in Australia
To:      Multiple recipients of list CHIPDIR-L
Subject: Re: speech IC*

Try the SP0256A-AL2 made be GI should be perfect for what you want.

Date:    19981228
From:    Chris Hostetler (Volt Computer)
To:      Jaap van Ganswijk
Subject: looking for voice synthesizer

On the page: local

You list a few voice synthesizers. I have found the SPO256-AL2 and even built a circuit for it and have it running fine but I have to make words by adding allophones together. I already emailed Ashok for info about obtaining the CTS256A-AL2.

But Im wondering about the other chips listed on that page. Can any of them recieve ASCII text and speak it, like the SPO256-AL2/CTS256A-AL2 combo? I tried finding more info on the other chips listed but the only page that any search engine finds is your page (I tried many search engines). Philips and Siemens web sites don't turn up any info either when searching on their local sites.

Any help you could give me on where to find them or data sheets (so I can tell if they would even work in my application) would be greatly appreciated!!! :)

Date:    19981228
From:    Jaap van Ganswijk
To:      Chris Hostetler (Volt Computer)
Subject: Re: looking for voice synthesizer

I really don't know much about it, except that this chip seems to remain
very popular although it's from about 1980 (I think) and hardly seems to
be available anymore.

If I had a voice synthesis problem to solve, I'd probably ask my local
distributors of OKI, ISD and the DSP* makers (AD(?), Motorola and TI)
for more information. It's hard to believe that 1980's chip technology
still isn't surpassed...

But this issue is dear to my heart for some reason, so I'd really
welcome more information about it, to put on my site. Could you, for
example send me a sample of the output in WAV format? It would
need to be as small as possible but still show the quality of the
generator. It might also be interesting to have besides an english text
a non-english text like german or french spoken.

Date:    19990101
From:    Chris (TheOne) Hostetler
To:      Jaap van Ganswijk
Subject: Re: Looking for voice synthesizer

Jaap, Just an FYI*...

I found what looks to be exactly what I need. You may want to add this to your site.

Its a complete circuit that converts ASCII text to speech. There are several models to choose from (many are designed to be connected to a PC). The one I'm looking at is designed to be connected to just about any microcotroller. It is a tad spendy at $150, but the amount of time and energy I have spent trying to get the SPO256-AL2 to perform the same function, it seems worth it for my purposes. Not to mention this thing does TONS more than I could ever do with the SPO256...

Anyway, check it out at . The model Im looking at it is the V860X (it has its own link in the main menu).

I emailed the company and asked for a wav file recording of them talking. If they send me one, I'll forward it on to you.

Doh! maybe I should have browsed the site a bit more first... :)

There is already a wav file of the speech produced by there products at:

It is about 600k I think (but pretty long also) , so you will have to cut it down, downsample it, convert it, or whatever you want. Or you can just link to it from your site I guess.

I have a wav file of the SPO256-AL2 already on my site (in my sig). If I find other chips and get them working, I'll have them posted to the same site along with a wav file.

Thanks for the info, I'll check around with TI and Motorola. I too find it hard to believe that no one has a better chip than from a chip from the 80's. :)
I'll let you know if I find anything.

FYI*, Ashok pointed me to for the CTS256A-AL2 chip, which they did have them for $4.99 and I was able to order a few....

Again, THANKS!


Date:    19990723
From:    Hugh Watkins
To:      Jaap van Ganswijk
Subject: Re: SPO256 for $5

I have done some further research on this and thought I would share what I have found.

Nobody that I can find has a supply of SPO256 chips. Everyone I have talked to said that bgmicro at was the last one to have a supply and they are out. The CTS256 chip is still available from jdr microdevices at

I guess moving forward there are several choices. The easiest is the V860X family from RC Systems at Parallel or Serial in and voice synthesis out. Doesn't get much easier than that. The downside on this solution is cost, between 150-190 dollars. I did see that they are going to sell the chip set directly, but there is no indication of how much these will cost.

One of the most promising I've seen is the the MSM7630 from oki at It looks like with this chip and a dictionary rom (OKI part MR27V3202-17 for US english... did I mention it's multi lingual??) and some ram* work area you are pretty much set. The Chip should be in the 10-20 dollar range and I'm not sure how much the dictionary rom is. This was developed with a company named Elan at I am trying to get some chips now to play around with, but not having much luck. It is a new product so supplies are scarce right now. If you know of a good OKI distributor let me know.

There are some other solutions like the voice recorders from ISD, but these will only have a limited vocabulary. There is also the Digitalker chip set available at Jameco, but again you have a limit vocabulary.

I really think that with PC's the need for dedicated chip sets has dried up. The good thing is it seems like we've come full circle where it's cheap enough to put the code in ROM and have a low cost processor do the work (i.e. the OKI chips).

See also

index.htm Index page for sound chips
melody.htm About melody generating chips
speechge.htm About speech/voice generation/synthesis
speechrc.htm About speech recording
voicegen.htm About voice/speech generation/synthesis
voicercg.htm About voice/speech recognition
../../oth/voicerec.txt FAQ about voice recognition processors

Ad for PCI video capturing boards by Dektec.
Goto: Main Mirror About Author
Register: Yourself Company
Feedback: Correction Addition Question
Request quote: Chips (Deutsch) Chips (English) Chips (Nederlands)

Advertisement by Adprov

Viewable with any browser

page modified: 20030711
page compiled: 20080306
site compiled:  Version GIF