Has anyone ever reverse engineered narrator.device?

[Thomas Richter]

Again, in case this wasn't clear: "narrator" is not abandoned. Softvoice exists as private entity, and their product exists, and you can buy it. You can even buy the narrator.device sources if you want to, except that - if you ask me - their price does not match its value. Narrator was a nice toy, but is not a necessity.

[dlfrsilver]

Originally Posted by Thomas Richter

Again, in case this wasn't clear: "narrator" is not abandoned. Softvoice exists as private entity, and their product exists, and you can buy it. You can even buy the narrator.device sources if you want to, except that - if you ask me - their price does not match its value. Narrator was a nice toy, but is not a necessity.

Why not trying to make a new program for this and then discard the old narrator ?

[Karlos]

Originally Posted by dlfrsilver

Why not trying to make a new program for this and then discard the old narrator ?

I suggested this. The main issue is that some people want only the original, and given the generally retro nature of the platform this is not an unreasonable wish. That basically leaves buying the legit sources or reverse engineering.

A clean room reimplementation of the original might be a fair compromise but I don't know how people feel about that.

[grond]

Originally Posted by Thomas Richter

No, it is "generally illegal" in an appropriate (legal) use of the word "general". It is only legal in certain exceptions (thus not "general"), and only if you want to establish an interface to an otherwise closed system.

Um, reverse-engineering generally *is* legal unless an exception applies. These exceptions come from copyright law, patent law, unfair competition law and probably some more. Examples for such an exception is direct immitation (often called "slavish imitation") rather than making use of the knowledge you gained (unfair competition law) or deriving a work from somebody else's work rather than making your own implementation (copyright law). Furthermore, what is not legal is circumvention of functioning copy-protection mechanisms.

The reason why reverse-engineering is legal is that the interest of the public to have access to products and services of best quality and lowest price, i.e. to have competition in the marketplace, overweighs the individuals' or companies' interest in their intellectual property.

[Thomas Richter]

So you call "copyright law" an "exception"? Interesting. The narrator.device is copyrighted, and its source code headers say so.

[grond]

Originally Posted by Thomas Richter

So you call "copyright law" an "exception"? Interesting. The narrator.device is copyrighted, and its source code headers say so.

Yes, the law is constructed like this. You can do anything you want unless it is explicitly prohibited. This is the nature of the constitutional state.

Copyright protection does not mean you can't reverse-engineer and learn about the inner workings of a piece of software. It means you can't derive work from copyrighted work without the copyright holder's permission. That is what a "cleanroom implementation" usually refers to: reverse-engineer to learn how something works and then use that knowledge to make your own implementation. That's perfectly legal.

[copse]

I looked at reverse engineering this many years ago, and also looked at all the alternative software projects Linux or otherwise. There were some good options. I wish I had my notes. I just wanted something lightweight that could do English voices and this was that.

Nowadays, we can also look to the "AI" to generate voices and even translate between languages. That's another option.

This thread is kind of a moot point. Someone who does not care about copyright might go ahead and do whatever they want. Those who do, might avoid using whatever is made or may never even see it :-)

[Bruce Abbott]

Originally Posted by Thomas Richter

Again, in case this wasn't clear: "narrator" is not abandoned. SoftVoice, Inc. exists as private entity, and their product exists, and you can buy it.

Narrator device is a product that they made exclusively for Commodore, the distribution rights to which are now owned by Cloanto. According the Amiga Forever Workbench 1.x-2.x Improvements webpage, they "Restored licensed speech-synthesis files to Devs, L, Libs, Utilities". If SoftVoice Inc. had a problem with that they could sue Cloanto. AFAIK that hasn't happened, which indicates that they are OK with it. Therefore anyone who buys WB 1.x/2.x disks from Cloanto today is getting a legal copy of the narrator device. In many countries this means they are permitted to disassemble the code and/or modify it for their own personal use.

You can even buy the narrator.device sources if you want to, except that - if you ask me - their price does not match its value.

We have the executable so we don't need the source code.

Narrator was a nice toy, but is not a necessity.

You could argue that the Amiga itself isn't essential, and you would be right. But if you want a full retro experience then it's essential. Some Amiga programs absolutely require it, eg. the OCS version of Word Construction Set.

Let's get real here. SoftVoice have effectively abandoned the Amiga. If they thought they could make any money out of it they would either continue to support the platform directly or license the source code at an attractive price. But they didn't, which means they think it's (currently) worthless.

But if people start making a fuss about reverse-engineering it even if we have the law on our side it could attract unwelcome attention from their lawyers. Nobody wants to be at the wrong end of a DCMA take-down notice or lawsuit - valid or not. So let's not talk seriously about resourcing code or publishing intimate details of how the device works, since that is the kind of thing they would jump on. Instead we will cower in the shadows hoping nobody notices what we are doing and spoils the fun.

The way these outfits come out of the woodwork to pick over the Amiga's bleached bones looking for something to monetize is disgusting. Sometimes I wish it had really died with Commodore, and all the assets made free like some other retro platforms have. If I was a billionaire I would make that happen. But since I am only a poor pensioner I will just have to continue the Amiga tradition of hacking code while ignoring the dire treats from legal eagles. Like Mr Robot, I hack everything - and there's nothing you can do about it!

[Thomas Richter]

Originally Posted by Bruce Abbott

Narrator device is a product that they made exclusively for Commodore,

No, it's a port of a softvoice product they made for many contemporary computer systems. A port also exists for the Mac, so it is not at all exclusive to the Amiga. The narrator.device is the successor of S.A.M., also from Softvoice, they made for the Atari 8 bit and the C64.

Originally Posted by Bruce Abbott

the distribution rights to which are now owned by Cloanto.

The narrator device was only licensed to CBM in binary form. I spoke to Mr. Katz about this, and he was astonished if not even afraid that CBM had the source code of it. A license existed for AmigaOs 1.3, but the 2.0 version was not licensed, and then removed by CBM from subsequent releases. CBM had the rights to distribute it, yes.

Originally Posted by Bruce Abbott

According the Amiga Forever Workbench 1.x-2.x Improvements webpage, they "Restored licensed speech-synthesis files to Devs, L, Libs, Utilities". If SoftVoice Inc. had a problem with that they could sue Cloanto.

Licensed to what? As part of 1.3 due to a license contract with CBM. In source code form, it was actually not licensed to CBM, but again, I talked to Softvoice, not you.

Originally Posted by Bruce Abbott

AFAIK that hasn't happened, which indicates that they are OK with it.

That only indicates that they haven't done anything, not that "everything is OK". That is not the same thing. As far as 1.3 is concerned, they are bound by contract.

Originally Posted by Bruce Abbott

In many countries this means they are permitted to disassemble the code and/or modify it for their own personal use.

Again, the problem starts if you redistribute such code because it is a derived work, derived from copyrighted code. So that buys you nothing.

Originally Posted by Bruce Abbott

We have the executable so we don't need the source code.

If that is enough for you, then that is enough.

Originally Posted by Bruce Abbott

You could argue that the Amiga itself isn't essential, and you would be right. But if you want a full retro experience then it's essential.

For what? I haven't used narrator except as a toy. "intuition" is essential because the system would not even boot without it. The narrator is a rather minor component nothing else depends on it. Actually, that 2.1 and following worked perfectly without it proved that it was never essential.

Originally Posted by Bruce Abbott

Let's get real here. SoftVoice have effectively abandoned the Amiga. If they thought they could make any money out of it they would either continue to support the platform directly or license the source code at an attractive price. But they didn't, which means they think it's (currently) worthless.

That is not quite the situation. They say they would want to license, but the price point is not very attractive, to put it this way. Anyhow, this is a free country, the web page of SoftVoice exists, there is a contact form, you can contact Mr. Katz there, and I got a reply, and that's the result. If you know better, well, write a letter and see what you get. Maybe you have success were others did not.

Originally Posted by Bruce Abbott

But if people start making a fuss about reverse-engineering it even if we have the law on our side

For doing what? For using it as part of 1.3, or for modifying it? Thus, if you have a licensed 1.3 copy, then go and use it. If you don't, you cannot. So it cannot be part of an updated AmigaOs distribution because it isn't licensed to include it there, quite simple.

Originally Posted by Bruce Abbott

it could attract unwelcome attention from their lawyers. Nobody wants to be at the wrong end of a DCMA take-down notice or lawsuit - valid or not. So let's not talk seriously about resourcing code or publishing intimate details of how the device works, since that is the kind of thing they would jump on. Instead we will cower in the shadows hoping nobody notices what we are doing and spoils the fun.

Look, again, Softvoice *is* aware of AmigaOs, the license price they ask for is just not attractive. They are aware because people from the development team already spoke to them. Actually, I already spoke to them, probably six years ago, and apparently somebody else afterwards.

Originally Posted by Bruce Abbott

The way these outfits come out of the woodwork to pick over the Amiga's bleached bones looking for something to monetize is disgusting.

That's not an "Amiga product", it is a "SoftVoice" product we talk about, and they can do with their product whatever they like.

Originally Posted by Bruce Abbott

Sometimes I wish it had really died with Commodore, and all the assets made free like some other retro platforms have.

"Made free" buy whom? Even if a company dies, it does not mean that the rights go away. They are assets of the company, and it is the job of the liquidator to montarize them to pay off the bills. Thus, "nothing becomes free", probably only in your dreams.

Originally Posted by Bruce Abbott

If I was a billionaire I would make that happen. But since I am only a poor pensioner I will just have to continue the Amiga tradition of hacking code while ignoring the dire treats from legal eagles. Like Mr Robot, I hack everything - and there's nothing you can do about it!

It's not my property, but SoftVoice could do something about it as soon as you distribute derived work. But that's a problem between you and them.

[Matt_H]

Originally Posted by Bruce Abbott

According the Amiga Forever Workbench 1.x-2.x Improvements webpage, they "Restored licensed speech-synthesis files to Devs, L, Libs, Utilities". If SoftVoice Inc. had a problem with that they could sue Cloanto. AFAIK that hasn't happened, which indicates that they are OK with it.

As I recall, Cloanto negotiated a new agreement with SoftVoice some years ago so that they could include narrator.device, etc. with Amiga Forever.

[Thorham]

Originally Posted by Thomas Richter

For what? I haven't used narrator except as a toy.

Text to speech isn't a toy. Remember Steven Hawking? There's also screen reading software for blind people, of course.

[Thomas Richter]

Did Steven Hawking use the narrator.device on the Amiga? Hardly. I'm talking about the usefulness of the narrator.device (and that implies the Amiga) - and which serious applications had it found? Like, for example, is it integrated into the Os? The console maybe? For blind people maybe? No.

This *IS* a toy - it was used (as on the Mac) as a tech-demo for the Amiga.

Anyhow, if you want to help to improve the situation and consider it important, either start collecting money to buy a license from Softvoice, or start porting another text to speech engine.

[Minuous]

Originally Posted by Nightfox

perhaps extend it do do some things such as output to file instead of just outputting straight to audio

http://aminet.net/package/mus/edit/Speak2Sample

[AmigaHope]

How much software bangs original content into narrator.device? A *lot* of it used translator.device. You could probably write a translator.device equivalent that spits out phonemes that are completely incompatible with the AmigaOS narrator.device and instead have your own narrator.device replacement that understands them and just pipes them into some other TTS engine, completely obviating the need for reverse engineering narrator.device.

Then you could have a fallback mode for narrator.device for any software that speaks phonemes directly, and just does a half-assed job of interpreting the phonemes. You could figure out a good phoneme table pretty quickly just by capturing output from translator.device and then feeding it into narrator.device and sampling the output, then deciding what each phoneme sounds like, until you have all the phonemes mapped.

No disassembly of narrator.device required.

[Locutus]

There's a bunch of open source TTS engines, some of them rather old and originally targeted at slow hardware in the same performance bracket as a =>030 Amiga (Mbrola was for VAX, eSpeak RISCOS ARM2/3).

If you really really would genuinely want this, building a shim layer on top of one of those that replicates the translator or narrator library interfaces shouldn't even be that hard (okay, famous last words).

SoftVoice has a website with samples up and even their latest version of the software sounds like generic 80's robo-voice, so you aren't going to be missing much with an alternative implementation.

I kind of wonder if the Softvoice guy has had any sales of his product in years, maybe he tries to target licensees needing TTS on super low power/performance platforms but even there it'd be a hard pill to swallow if we are talking such big license fees. ...But that's neither here or there.

I think the fact nobody has done this after more than 25 years might give a hint at how much weight this feature actually carries.

[nogginthenog]

There's some documentation on the phonemes here:

https://archive.org/details/Amiga_BA.../n298/mode/1up

[Dunny]

As some here may know, I maintain a BASIC interpreter. One thing I would love to add is a SAY command - and a detailed description of the algorithms employed, how the sound is generated etc etc so that I can do that would be amazing, especially if it sounds authentically like an Amiga.

[Olaf Barthel]

Originally Posted by nogginthenog

There's some documentation on the phonemes here:

https://archive.org/details/Amiga_BA.../n298/mode/1up

This documentation references the "Arpabet", which is likely designed for English language speakers. As the means to pronounce words in a different language from English this may have merit, but just as with the capabilities of narrator.device, the scope of phonemes is limited. As Thomas already mentioned, the phonemes of a German "Umlaut" do not exist in English (but they certainly do in French, for example).

SoftVoice supporting Spanish makes good sense. Counted by the number of native speakers, Spanish is in the second place, after Chinese. English is in the third place.

[Dunny]

Originally Posted by nogginthenog

There's some documentation on the phonemes here:

https://archive.org/details/Amiga_BA.../n298/mode/1up

That's nice but what I need to know is how translator lib converts from plain english to phonemes, and how phonemes are built, modified for stress etc and played. Are they samples? Are they created on the fly? If so, how? etc etc.

[Thomas Richter]

Originally Posted by Dunny

That's nice but what I need to know is how translator lib converts from plain english to phonemes, and how phonemes are built, modified for stress etc and played. Are they samples? Are they created on the fly? If so, how? etc etc.

Use the source, Luke!



http://aminet.net/package/util/libs/translator42


http://aminet.net/package/dev/src/trans42src


Essentially, it is a big list of rules, along with an even bigger list of exceptions, by word. I remember that I played a lot with it attempting to create a somewhat better version of Geman translator rules - without really succeeding.



Unfortunately, a lot of the melody of speech are built into the narrator.device, and thus not accessible on the level of the translator. This is why every language sounds as if an american attempts to speak a foreign language (or, shortly, "not convincing").